ARC mm updates: support 3/4 levels and asm-generic/pgalloc
Subsystems (2)
Vineet Gupta
maintainerv2 cover ARC mm updates: support 3/4 levels and asm-generic/pgalloc
Hi,
Big pile of ARC mm changes to prepare for 3 or 4 levels of paging (from current 2) needed for new hardware page walked MMUv6 (in aRCv3 ISA based cores).
Most of these changes are incremental cleanups to make way for 14/18 and 15/18 which actually imeplement the new levels (in existing ARCv2 port) and worth a critical eye.
CC'ing some of you guys dealing with page tables for a while :-) to spot any obvious gotchas.
Thx,
-Vineet
Changes since v1 [1]
- Switched ARC to asm-generic/pgalloc.h (so struct page based pgtable_t) [Mike Rapoport]
- Dropped {pud,pmd}_alloc_one/{pud,pmd}_free provided by asm-generic/pgalloc.h [Mike Rapoport]
- Negative diffstat now due to above
- Added BUILD_BUG_ON() to arch/arc/mm/init.c for sanity of table sizes
- Consolidated 2 patches related to ARC_USE_SCRATCH_REG [Mike Rapoport]
- Reworked how mmu is re-enabled in entry code [Jose Abreu]
[1] http://lists.infradead.org/pipermail/linux-snps-arc/2021-August/005326.html
Vineet Gupta (19):
ARC: mm: use SCRATCH_DATA0 register for caching pgdir in ARCv2 only
ARC: mm: remove tlb paranoid code
ARC: mm: move mmu/cache externs out to setup.h
ARC: mm: Fixes to allow STRICT_MM_TYPECHECKS
ARC: mm: Enable STRICT_MM_TYPECHECKS
ARC: ioremap: use more commonly used PAGE_KERNEL based uncached flag
ARC: mm: pmd_populate* to use the canonical set_pmd (and drop pmd_set)
ARC: mm: switch pgtable_t back to struct page *
ARC: mm: switch to asm-generic/pgalloc.h
ARC: mm: non-functional code cleanup ahead of 3 levels
ARC: mm: move MMU specific bits out of ASID allocator
ARC: mm: move MMU specific bits out of entry code ...
ARC: mm: disintegrate mmu.h (arcv2 bits out)
ARC: mm: disintegrate pgtable.h into levels and flags
ARC: mm: hack to allow 2 level build with 4 level code
ARC: mm: support 3 levels of page tables
ARC: mm: support 4 levels of page tables
ARC: mm: vmalloc sync from kernel to user table to update PMD ...
ARC: mm: introduce _PAGE_TABLE to explicitly link pgd,pud,pmd entries
arch/arc/Kconfig | 7 +- arch/arc/include/asm/cache.h | 4 - arch/arc/include/asm/entry-compact.h | 8 - arch/arc/include/asm/mmu-arcv2.h | 103 +++++++ arch/arc/include/asm/mmu.h | 73 +---- arch/arc/include/asm/mmu_context.h | 28 +- arch/arc/include/asm/page.h | 74 +++-- arch/arc/include/asm/pgalloc.h | 81 ++---- arch/arc/include/asm/pgtable-bits-arcv2.h | 151 +++++++++++ arch/arc/include/asm/pgtable-levels.h | 179 ++++++++++++ arch/arc/include/asm/pgtable.h | 315 +--------------------- arch/arc/include/asm/processor.h | 2 +- arch/arc/include/asm/setup.h | 12 +- arch/arc/kernel/entry-arcv2.S | 1 + arch/arc/kernel/entry.S | 7 +- arch/arc/mm/fault.c | 20 +- arch/arc/mm/init.c | 5 + arch/arc/mm/ioremap.c | 3 +- arch/arc/mm/tlb.c | 68 +---- arch/arc/mm/tlbex.S | 78 ++---- 20 files changed, 591 insertions(+), 628 deletions(-) create mode 100644 arch/arc/include/asm/mmu-arcv2.h create mode 100644 arch/arc/include/asm/pgtable-bits-arcv2.h create mode 100644 arch/arc/include/asm/pgtable-levels.h
Signature · 1 line
2.25.1
Vineet Gupta
maintainerPATCH v2 1/19 ARC: mm: use SCRATCH_DATA0 register for caching pgdir in ARCv2 only
MMU SCRATCH_DATA0 register is intended to cache task pgd. However in ARC700 SMP port, it has to be repurposed for reentrant interrupt handling, while UP port doesn't. We currently ahandle boe usecases using a fabricated which has usual issues of dependency nesting and ugliness.
So clean this up: for ARC700 don't use to cache pgd (even in UP) and do the opposite for ARCv2.
And while here, switch to canonical pgd_offset().
--- arch/arc/include/asm/entry-compact.h | 8 -------- arch/arc/include/asm/mmu.h | 4 ---- arch/arc/include/asm/mmu_context.h | 2 +- arch/arc/include/asm/pgtable.h | 23 ----------------------- arch/arc/mm/fault.c | 2 +- arch/arc/mm/tlb.c | 4 ++-- arch/arc/mm/tlbex.S | 2 +- 7 files changed, 5 insertions(+), 40 deletions(-)
arch/arc/include/asm/entry-compact.h +0 -8
@@ -126,19 +126,11 @@* to be saved again on kernel mode stack, as part of pt_regs. *-------------------------------------------------------------*/ .macro PROLOG_FREEUP_REG reg, mem -#ifndef ARC_USE_SCRATCH_REG - sr \reg, [ARC_REG_SCRATCH_DATA0] -#else st \reg, [\mem] -#endif .endm .macro PROLOG_RESTORE_REG reg, mem -#ifndef ARC_USE_SCRATCH_REG - lr \reg, [ARC_REG_SCRATCH_DATA0] -#else ld \reg, [\mem] -#endif .endm/*--------------------------------------------------------------
arch/arc/include/asm/mmu.h +0 -4
@@ -31,10 +31,6 @@#define ARC_REG_SCRATCH_DATA0 0x46c#endif -#if defined(CONFIG_ISA_ARCV2 ) || !defined(CONFIG_SMP ) -#define ARC_USE_SCRATCH_REG -#endif -/* Bits in MMU PID register */ #define __TLB_ENABLE (1 << 31)#define __PROG_ENABLE (1 << 30)
arch/arc/include/asm/mmu_context.h +1 -1
@@ -146,7 +146,7 @@ static inline void switch_mm(struct mm_struct *prev, struct mm_struct *next,*/ cpumask_set_cpu (cpu,mm_cpumask (next)); -#ifdef ARC_USE_SCRATCH_REG +#ifdef CONFIG_ISA_ARCV2 /* PGD cached in MMU reg to avoid 3 mem lookups: task->mm->pgd */ write_aux_reg (ARC_REG_SCRATCH_DATA0 , next->pgd);#endif
arch/arc/include/asm/pgtable.h +0 -23
@@ -284,29 +284,6 @@ static inline void set_pte_at(struct mm_struct *mm, unsigned long addr,set_pte (ptep, pteval); } -/* -* Macro to quickly access the PGD entry, utlising the fact that some -* arch may cache the pointer to Page Directory of "current" task -* in a MMU register -* -* Thus task->mm->pgd (3 pointer dereferences, cache misses etc simply -* becomes read a register -* -* ********CAUTION*******: -* Kernel code might be dealing with some mm_struct of NON "current" -* Thus use this macro only when you are certain that "current" is current -* e.g. when dealing with signal frame setup code etc -*/ -#ifdef ARC_USE_SCRATCH_REG -#define pgd_offset_fast (mm, addr) \ -({ \ - pgd_t *pgd_base = (pgd_t *) read_aux_reg(ARC_REG_SCRATCH_DATA0); \ - pgd_base + pgd_index(addr); \ -}) -#else -#define pgd_offset_fast (mm, addr) pgd_offset(mm, addr) -#endif -extern pgd_t swapper_pg_dir[]__aligned(PAGE_SIZE) ;void update_mmu_cache (struct vm_area_struct *vma,unsigned long address,pte_t *ptep);
arch/arc/mm/fault.c +1 -1
@@ -33,7 +33,7 @@ noinline static int handle_kernel_vaddr_fault(unsigned long address)pud_t *pud, *pud_k;pmd_t *pmd, *pmd_k; - pgd =pgd_offset_fast (current->active_mm, address); + pgd =pgd_offset (current->active_mm, address); pgd_k =pgd_offset_k (address);if (!pgd_present (*pgd_k))
arch/arc/mm/tlb.c +2 -2
@@ -719,8 +719,8 @@ void arc_mmu_init(void)/* Enable the MMU */ write_aux_reg (ARC_REG_PID ,MMU_ENABLE ); -/* In smp we use this reg for interrupt 1 scratch */ -#ifdef ARC_USE_SCRATCH_REG +/* In arc700/smp needed for re-entrant interrupt handling */ +#ifdef CONFIG_ISA_ARCV2 /* swapper_pg_dir is the pgd for the kernel, used by vmalloc */ write_aux_reg (ARC_REG_SCRATCH_DATA0 , swapper_pg_dir);#endif
arch/arc/mm/tlbex.S +1 -1
@@ -202,7 +202,7 @@ ex_saved_reg1: lr r2, [efa] -#ifdef ARC_USE_SCRATCH_REG +#ifdef CONFIG_ISA_ARCV2 lr r1, [ARC_REG_SCRATCH_DATA0] ; current pgd #else GET_CURR_TASK_ON_CPU r1
Signature · 1 line
2.25.1
Re: [PATCH v2 01/19] ARC: mm: use SCRATCH_DATA0 register for caching pgdir in ARCv2 only
Quoted 3 lines · Vineet Gupta
MMU
SCRATCH_DATA0register is intended to cache task pgd. However in ARC700 SMP port, it has to be repurposed for reentrant interrupt handling, while UP port doesn't. We currently ahandle boe usecases
^ handle both
maybe ':set spell' for changelog editing? ;-)
Quoted 149 lines
using a fabricated which has usual issues of dependency nesting and ugliness.
So clean this up: for ARC700 don't use to cache pgd (even in UP) and do the opposite for ARCv2.
And while here, switch to canonical
pgd_offset().Signed-off-by Vineet Gupta <[email protected]>--- arch/arc/include/asm/entry-compact.h | 8 -------- arch/arc/include/asm/mmu.h | 4 ---- arch/arc/include/asm/mmu_context.h | 2 +- arch/arc/include/asm/pgtable.h | 23 ----------------------- arch/arc/mm/fault.c | 2 +- arch/arc/mm/tlb.c | 4 ++-- arch/arc/mm/tlbex.S | 2 +- 7 files changed, 5 insertions(+), 40 deletions(-)arch/arc/include/asm/entry-compact.h +0
-8@@ -126,19 +126,11 @@* to be saved again on kernel mode stack, as part of pt_regs. *-------------------------------------------------------------*/ .macro PROLOG_FREEUP_REG reg, mem -#ifndef ARC_USE_SCRATCH_REG - sr \reg, [ARC_REG_SCRATCH_DATA0] -#else st \reg, [\mem] -#endif .endm .macro PROLOG_RESTORE_REG reg, mem -#ifndef ARC_USE_SCRATCH_REG - lr \reg, [ARC_REG_SCRATCH_DATA0] -#else ld \reg, [\mem] -#endif .endm/*-------------------------------------------------------------- arch/arc/include/asm/mmu.h +0
-4@@ -31,10 +31,6 @@#define ARC_REG_SCRATCH_DATA0 0x46c#endif -#if defined(CONFIG_ISA_ARCV2 ) || !defined(CONFIG_SMP ) -#define ARC_USE_SCRATCH_REG -#endif -/* Bits in MMU PID register */ #define __TLB_ENABLE (1 << 31)#define __PROG_ENABLE (1 << 30)arch/arc/include/asm/mmu_context.h +1
-1@@ -146,7 +146,7 @@ static inline void switch_mm(struct mm_struct *prev, struct mm_struct *next,*/ cpumask_set_cpu (cpu,mm_cpumask (next)); -#ifdef ARC_USE_SCRATCH_REG +#ifdef CONFIG_ISA_ARCV2 /* PGD cached in MMU reg to avoid 3 mem lookups: task->mm->pgd */ write_aux_reg (ARC_REG_SCRATCH_DATA0 , next->pgd);#endif arch/arc/include/asm/pgtable.h +0
-23@@ -284,29 +284,6 @@ static inline void set_pte_at(struct mm_struct *mm, unsigned long addr,set_pte (ptep, pteval); } -/* -* Macro to quickly access the PGD entry, utlising the fact that some -* arch may cache the pointer to Page Directory of "current" task -* in a MMU register -* -* Thus task->mm->pgd (3 pointer dereferences, cache misses etc simply -* becomes read a register -* -* ********CAUTION*******: -* Kernel code might be dealing with some mm_struct of NON "current" -* Thus use this macro only when you are certain that "current" is current -* e.g. when dealing with signal frame setup code etc -*/ -#ifdef ARC_USE_SCRATCH_REG -#define pgd_offset_fast (mm, addr) \ -({ \ - pgd_t *pgd_base = (pgd_t *) read_aux_reg(ARC_REG_SCRATCH_DATA0); \ - pgd_base + pgd_index(addr); \ -}) -#else -#define pgd_offset_fast (mm, addr) pgd_offset(mm, addr) -#endif -extern pgd_t swapper_pg_dir[]__aligned(PAGE_SIZE) ;void update_mmu_cache (struct vm_area_struct *vma,unsigned long address,pte_t *ptep);arch/arc/mm/fault.c +1
-1@@ -33,7 +33,7 @@ noinline static int handle_kernel_vaddr_fault(unsigned long address)pud_t *pud, *pud_k;pmd_t *pmd, *pmd_k; - pgd =pgd_offset_fast (current->active_mm, address); + pgd =pgd_offset (current->active_mm, address); pgd_k =pgd_offset_k (address);if (!pgd_present (*pgd_k))arch/arc/mm/tlb.c +2
-2@@ -719,8 +719,8 @@ void arc_mmu_init(void)/* Enable the MMU */ write_aux_reg (ARC_REG_PID ,MMU_ENABLE ); -/* In smp we use this reg for interrupt 1 scratch */ -#ifdef ARC_USE_SCRATCH_REG +/* In arc700/smp needed for re-entrant interrupt handling */ +#ifdef CONFIG_ISA_ARCV2 /* swapper_pg_dir is the pgd for the kernel, used by vmalloc */ write_aux_reg (ARC_REG_SCRATCH_DATA0 , swapper_pg_dir);#endif arch/arc/mm/tlbex.S +1
-1@@ -202,7 +202,7 @@ ex_saved_reg1: lr r2, [efa] -#ifdef ARC_USE_SCRATCH_REG +#ifdef CONFIG_ISA_ARCV2 lr r1, [ARC_REG_SCRATCH_DATA0] ; current pgd #else GET_CURR_TASK_ON_CPU r1Signature · 1 line
2.25.1
Signature · 2 lines
Sincerely yours, Mike.
Vineet Gupta
maintainerRe: [PATCH v2 01/19] ARC: mm: use SCRATCH_DATA0 register for caching pgdir in ARCv2 only
Quoted 7 lines · Mike Rapoport
Quoted 3 lines · Vineet Gupta
MMU
SCRATCH_DATA0register is intended to cache task pgd. However in ARC700 SMP port, it has to be repurposed for reentrant interrupt handling, while UP port doesn't. We currently ahandle boe usecases^ handle both
maybe ':set spell' for changelog editing? ;-)
Seriously, about time I stopped fat-fingering
Vineet Gupta
maintainerPATCH v2 2/19 ARC: mm: remove tlb paranoid code
In reply to Vineet GuptaThis was used back in arc700 days when ASID allocator was fragile. Not needed in last 5 years
--- arch/arc/Kconfig | 3 --- arch/arc/include/asm/mmu.h | 6 ----- arch/arc/mm/tlb.c | 40 ------------------------------ arch/arc/mm/tlbex.S | 50 -------------------------------------- 4 files changed, 99 deletions(-)
arch/arc/Kconfig +0 -3
@@ -537,9 +537,6 @@ config ARC_DW2_UNWIND If you don't debug the kernel, you can say N, but we may not be able to solve problems without frame unwind information -config ARC_DBG_TLB_PARANOIA - bool "Paranoia Checks in Low Level TLB Handlers" - config ARC_DBG_JUMP_LABEL bool "Paranoid checks in Static Keys (jump labels) code" depends on JUMP_LABEL
arch/arc/include/asm/mmu.h +0 -6
@@ -64,12 +64,6 @@ typedef struct {unsigned long asid[NR_CPUS ];/* 8 bit MMU PID + Generation cycle */ } mm_context_t; -#ifdef CONFIG_ARC_DBG_TLB_PARANOIA -void tlb_paranoid_check (unsigned int mm_asid,unsigned long address); -#else -#define tlb_paranoid_check (a, b) -#endif -void arc_mmu_init (void );extern char *arc_mmu_mumbojumbo (int cpu_id,char *buf,int len);void read_decode_mmu_bcr (void );
arch/arc/mm/tlb.c +0 -40
@@ -400,7 +400,6 @@ void create_tlb(struct vm_area_struct *vma, unsigned long vaddr, pte_t *ptep)* * Removing the assumption involves * -Using vma->mm->context{ASID,SASID}, as opposed to MMU reg. -* -Fix the TLB paranoid debug code to not trigger false negatives. * -More importantly it makes this handler inconsistent with fast-path * TLB Refill handler which always deals with "current" *
@@ -423,8 +422,6 @@ void create_tlb(struct vm_area_struct *vma, unsigned long vaddr, pte_t *ptep)local_irq_save (flags); -tlb_paranoid_check (asid_mm (vma->vm_mm,smp_processor_id ()), vaddr); - vaddr &=PAGE_MASK ;/* update this PTE credentials */
@@ -818,40 +815,3 @@ void do_tlb_overlap_fault(unsigned long cause, unsigned long address,local_irq_restore (flags); } - -/*********************************************************************** -* Diagnostic Routines -* -Called from Low Level TLB Handlers if things don;t look good -**********************************************************************/ - -#ifdef CONFIG_ARC_DBG_TLB_PARANOIA - -/* -* Low Level ASM TLB handler calls this if it finds that HW and SW ASIDS -* don't match -*/ -void print_asid_mismatch (int mm_asid,int mmu_asid,int is_fast_path) -{ -pr_emerg ("ASID Mismatch in %s Path Handler: sw-pid=0x%x hw-pid=0x%x\n" , - is_fast_path ?"Fast" :"Slow" , mm_asid, mmu_asid); - - __asm__ __volatile__("flag 1" ); -} - -void tlb_paranoid_check (unsigned int mm_asid,unsigned long addr) -{ -unsigned int mmu_asid; - - mmu_asid =read_aux_reg (ARC_REG_PID ) &0xff ; - -/* -* At the time of a TLB miss/installation -* - HW version needs to match SW version -* - SW needs to have a valid ASID -*/ -if (addr <0x70000000 && - ((mm_asid ==MM_CTXT_NO_ASID ) || - (mmu_asid != (mm_asid &MM_CTXT_ASID_MASK )))) -print_asid_mismatch (mm_asid, mmu_asid,0 ); -} -#endif
arch/arc/mm/tlbex.S +0 -50
62 lines not on this page · raw message
Signature · 1 line
2.25.1
Vineet Gupta
maintainerPATCH v2 3/19 ARC: mm: move mmu/cache externs out to setup.h
In reply to Vineet Gupta--- arch/arc/include/asm/cache.h | 4 ---- arch/arc/include/asm/mmu.h | 4 ---- arch/arc/include/asm/setup.h | 12 ++++++++++-- 3 files changed, 10 insertions(+), 10 deletions(-)
arch/arc/include/asm/cache.h +0 -4
@@ -62,10 +62,6 @@#define ARCH_SLAB_MINALIGN 8#endif -extern void arc_cache_init (void ); -extern char *arc_cache_mumbojumbo (int cpu_id,char *buf,int len); -extern void read_decode_cache_bcr (void ); -extern int ioc_enable;extern unsigned long perip_base, perip_end;
arch/arc/include/asm/mmu.h +0 -4
@@ -64,10 +64,6 @@ typedef struct {unsigned long asid[NR_CPUS ];/* 8 bit MMU PID + Generation cycle */ } mm_context_t; -void arc_mmu_init (void ); -extern char *arc_mmu_mumbojumbo (int cpu_id,char *buf,int len); -void read_decode_mmu_bcr (void ); -static inline int is_pae40_enabled (void ) {return IS_ENABLED (CONFIG_ARC_HAS_PAE40 );
arch/arc/include/asm/setup.h +10 -2
@@ -2,8 +2,8 @@/* * Copyright (C) 2004, 2007-2010, 2011-2012 Synopsys, Inc. (www.synopsys.com) */ -#ifndef __ASMARC_SETUP_H -#define __ASMARC_SETUP_H +#ifndef __ASM_ARC_SETUP_H +#define __ASM_ARC_SETUP_H#include <linux/types.h>
@@ -34,4 +34,12 @@ long __init arc_get_mem_sz(void);#define IS_AVAIL2 (v, s, cfg) IS_AVAIL1(v, s), IS_AVAIL1(v, IS_USED_CFG(cfg))#define IS_AVAIL3 (v, v2, s) IS_AVAIL1(v, s), IS_AVAIL1(v, IS_DISABLED_RUN(v2)) +extern void arc_mmu_init (void ); +extern char *arc_mmu_mumbojumbo (int cpu_id,char *buf,int len); +extern void read_decode_mmu_bcr (void ); + +extern void arc_cache_init (void ); +extern char *arc_cache_mumbojumbo (int cpu_id,char *buf,int len); +extern void read_decode_cache_bcr (void ); +#endif /* __ASMARC_SETUP_H */
Signature · 1 line
2.25.1
Re: [PATCH v2 03/19] ARC: mm: move mmu/cache externs out to setup.h
Heh,
"Don't pollute mmu.h and cache.h with some of ARC internal bootlog/setup related functions.
move them aside to setup.h"
is still not there :)
Quoted 68 lines · Vineet Gupta
Signed-off-by Vineet Gupta <[email protected]>--- arch/arc/include/asm/cache.h | 4 ---- arch/arc/include/asm/mmu.h | 4 ---- arch/arc/include/asm/setup.h | 12 ++++++++++-- 3 files changed, 10 insertions(+), 10 deletions(-)arch/arc/include/asm/cache.h +0
-4@@ -62,10 +62,6 @@#define ARCH_SLAB_MINALIGN 8#endif -extern void arc_cache_init (void ); -extern char *arc_cache_mumbojumbo (int cpu_id,char *buf,int len); -extern void read_decode_cache_bcr (void ); -extern int ioc_enable;extern unsigned long perip_base, perip_end;arch/arc/include/asm/mmu.h +0
-4@@ -64,10 +64,6 @@ typedef struct {unsigned long asid[NR_CPUS ];/* 8 bit MMU PID + Generation cycle */ } mm_context_t; -void arc_mmu_init (void ); -extern char *arc_mmu_mumbojumbo (int cpu_id,char *buf,int len); -void read_decode_mmu_bcr (void ); -static inline int is_pae40_enabled (void ) {return IS_ENABLED (CONFIG_ARC_HAS_PAE40 );arch/arc/include/asm/setup.h +10
-2@@ -2,8 +2,8 @@/* * Copyright (C) 2004, 2007-2010, 2011-2012 Synopsys, Inc. (www.synopsys.com) */ -#ifndef __ASMARC_SETUP_H -#define __ASMARC_SETUP_H +#ifndef __ASM_ARC_SETUP_H +#define __ASM_ARC_SETUP_H#include <linux/types.h> @@ -34,4 +34,12 @@ long __init arc_get_mem_sz(void);#define IS_AVAIL2 (v, s, cfg) IS_AVAIL1(v, s), IS_AVAIL1(v, IS_USED_CFG(cfg))#define IS_AVAIL3 (v, v2, s) IS_AVAIL1(v, s), IS_AVAIL1(v, IS_DISABLED_RUN(v2)) +extern void arc_mmu_init (void ); +extern char *arc_mmu_mumbojumbo (int cpu_id,char *buf,int len); +extern void read_decode_mmu_bcr (void ); + +extern void arc_cache_init (void ); +extern char *arc_cache_mumbojumbo (int cpu_id,char *buf,int len); +extern void read_decode_cache_bcr (void ); +#endif /* __ASMARC_SETUP_H */ Signature · 1 line
2.25.1
Signature · 2 lines
Sincerely yours, Mike.
Re: [PATCH v2 03/19] ARC: mm: move mmu/cache externs out to setup.h
Quoted 7 lines · Mike Rapoport
Heh,
"Don't pollute
mmu.handcache.hwith some of ARC internal bootlog/setup related functions.
move them aside tosetup.h"is still not there :)
Oops my bad, indeed missed out on that one. Fixed now.
Thx,
-Vineet
Vineet Gupta
maintainerPATCH v2 4/19 ARC: mm: Fixes to allow STRICT_MM_TYPECHECKS
In reply to Vineet Gupta--- arch/arc/mm/tlb.c | 13 ++++++++----- 1 file changed, 8 insertions(+), 5 deletions(-)
arch/arc/mm/tlb.c +8 -5
@@ -71,7 +71,7 @@ static void tlb_entry_erase(unsigned int vaddr_n_asid) } } -static void tlb_entry_insert (unsigned int pd0,pte_t pd1) +static void tlb_entry_insert (unsigned int pd0,phys_addr_t pd1) {unsigned int idx;
@@ -109,13 +109,16 @@ static void tlb_entry_erase(unsigned int vaddr_n_asid)write_aux_reg (ARC_REG_TLBCOMMAND , TLBDeleteEntry); } -static void tlb_entry_insert (unsigned int pd0,pte_t pd1) +static void tlb_entry_insert (unsigned int pd0,phys_addr_t pd1) {write_aux_reg (ARC_REG_TLBPD0 , pd0); -write_aux_reg (ARC_REG_TLBPD1 , pd1); -if (is_pae40_enabled ()) +if (!is_pae40_enabled ()) { +write_aux_reg (ARC_REG_TLBPD1 , pd1); + }else { +write_aux_reg (ARC_REG_TLBPD1 , pd1 &0xFFFFFFFF );write_aux_reg (ARC_REG_TLBPD1HI , (u64 )pd1 >>32 ); + }write_aux_reg (ARC_REG_TLBCOMMAND , TLBInsertEntry); }
@@ -391,7 +394,7 @@ void create_tlb(struct vm_area_struct *vma, unsigned long vaddr, pte_t *ptep)unsigned long flags;unsigned int asid_or_sasid, rwx;unsigned long pd0; -pte_t pd1; +phys_addr_t pd1;/* * create_tlb() assumes that current->mm == vma->mm, since
Signature · 7 lines
2.25.1 _______________________________________________ linux-snps-arc mailing list [email protected] http://lists.infradead.org/mailman/listinfo/linux-snps-arc
Vineet Gupta
maintainerPATCH v2 5/19 ARC: mm: Enable STRICT_MM_TYPECHECKS
In reply to Vineet GuptaIn the past I've refrained from doing this (at least 2 times) due to the slight code bloat due to ABI implications of pte_t etc becoming struct
Per ARC ABI, functions return struct via memory and not through register r0, even if the struct would fit in register(s)
- caller allocates space on stack and passes the address as first arg (r0), shifting rest of args by one
- callee creates return struct in memory (referenced via r0)
This time around the code actually shrunk slightly (due to subtle inlining heuristic effects), but still slightly inefficient due to return values passed through memory. That however seems like a small cost compared to maintenance burden given the impending new mmu support for page walk etc
--- arch/arc/include/asm/page.h | 26 -------------------------- arch/arc/mm/ioremap.c | 2 +- 2 files changed, 1 insertion(+), 27 deletions(-)
arch/arc/include/asm/page.h +0 -26
@@ -34,12 +34,6 @@ void copy_user_highpage(struct page *to, struct page *from, unsigned long u_vaddr, struct vm_area_struct *vma);void clear_user_page (void *to,unsigned long u_vaddr,struct page *page); -#undef STRICT_MM_TYPECHECKS - -#ifdef STRICT_MM_TYPECHECKS -/* - * These are used to make use of C type-checking.. - */ typedef struct { #ifdef CONFIG_ARC_HAS_PAE40 unsigned long long pte;
@@ -64,26 +58,6 @@ typedef struct {#define pte_pgprot (x) __pgprot(pte_val(x)) -#else /* !STRICT_MM_TYPECHECKS */ - -#ifdef CONFIG_ARC_HAS_PAE40 -typedef unsigned long long pte_t ; -#else -typedef unsigned long pte_t ; -#endif -typedef unsigned long pgd_t ; -typedef unsigned long pgprot_t ; - -#define pte_val (x) (x) -#define pgd_val (x) (x) -#define pgprot_val (x) (x) -#define __pte (x) (x) -#define __pgd (x) (x) -#define __pgprot (x) (x) -#define pte_pgprot (x) (x) - -#endif -typedef pte_t *pgtable_t ;/*
arch/arc/mm/ioremap.c +1 -1
@@ -39,7 +39,7 @@ void __iomem *ioremap(phys_addr_t paddr, unsigned long size)if (arc_uncached_addr_space (paddr))return (void __iomem *)(u32 )paddr; -return ioremap_prot (paddr, size,PAGE_KERNEL_NO_CACHE ); +return ioremap_prot (paddr, size,pgprot_val (PAGE_KERNEL_NO_CACHE )); }EXPORT_SYMBOL (ioremap);
Signature · 7 lines
2.25.1 _______________________________________________ linux-snps-arc mailing list [email protected] http://lists.infradead.org/mailman/listinfo/linux-snps-arc
Vineet Gupta
maintainerPATCH v2 6/19 ARC: ioremap: use more commonly used PAGE_KERNEL based uncached flag
In reply to Vineet Guptaand remove the one off uncached definition for ARC
--- arch/arc/include/asm/pgtable.h | 3 --- arch/arc/mm/ioremap.c | 3 ++- 2 files changed, 2 insertions(+), 4 deletions(-)
arch/arc/include/asm/pgtable.h +0 -3
@@ -103,9 +103,6 @@*/ #define PAGE_KERNEL __pgprot(_K_PAGE_PERMS | _PAGE_CACHEABLE) -/* ioremap */ -#define PAGE_KERNEL_NO_CACHE __pgprot(_K_PAGE_PERMS) -/* Masks for actual TLB "PD"s */ #define PTE_BITS_IN_PD0 (_PAGE_GLOBAL | _PAGE_PRESENT | _PAGE_HW_SZ)#define PTE_BITS_RWX (_PAGE_EXECUTE | _PAGE_WRITE | _PAGE_READ)
arch/arc/mm/ioremap.c +2 -1
@@ -39,7 +39,8 @@ void __iomem *ioremap(phys_addr_t paddr, unsigned long size)if (arc_uncached_addr_space (paddr))return (void __iomem *)(u32 )paddr; -return ioremap_prot (paddr, size,pgprot_val (PAGE_KERNEL_NO_CACHE )); +return ioremap_prot (paddr, size, +pgprot_val (pgprot_noncached (PAGE_KERNEL ))); }EXPORT_SYMBOL (ioremap);
Signature · 7 lines
2.25.1 _______________________________________________ linux-snps-arc mailing list [email protected] http://lists.infradead.org/mailman/listinfo/linux-snps-arc
Vineet Gupta
maintainerPATCH v2 7/19 ARC: mm: pmd_populate* to use the canonical set_pmd (and drop pmd_set)
In reply to Vineet Gupta--- arch/arc/include/asm/pgalloc.h | 14 ++++++++++---- arch/arc/include/asm/pgtable.h | 6 ------ 2 files changed, 10 insertions(+), 10 deletions(-)
arch/arc/include/asm/pgalloc.h +10 -4
@@ -35,13 +35,19 @@static inline void pmd_populate_kernel (struct mm_struct *mm,pmd_t *pmd,pte_t *pte) { -pmd_set (pmd, pte); +/* +* The cast to long below is OK in 32-bit PAE40 regime with long long pte +* Despite "wider" pte, the pte table needs to be in non-PAE low memory +* as all higher levels can only hold long pointers. +* +* The cast itself is needed given simplistic definition of set_pmd() +*/ +set_pmd (pmd,__pmd ((unsigned long )pte)); } -static inline void -pmd_populate (struct mm_struct *mm,pmd_t *pmd,pgtable_t ptep) +static inline void pmd_populate (struct mm_struct *mm,pmd_t *pmd,pgtable_t pte) { -pmd_set (pmd, (pte_t *) ptep); +set_pmd (pmd,__pmd ((unsigned long )pte)); }static inline int __get_order_pgd (void )
arch/arc/include/asm/pgtable.h +0 -6
@@ -222,12 +222,6 @@ extern char empty_zero_page[PAGE_SIZE];/* find the logical addr (phy for ARC) of the Page Tbl ref by PMD entry */ #define pmd_page_vaddr (pmd) (pmd_val(pmd) & PAGE_MASK) -/* In a 2 level sys, setup the PGD entry with PTE value */ -static inline void pmd_set (pmd_t *pmdp,pte_t *ptep) -{ -pmd_val (*pmdp) = (unsigned long )ptep; -} -#define pte_none (x) (!pte_val(x))#define pte_present (x) (pte_val(x) & _PAGE_PRESENT)#define pte_clear (mm, addr, ptep) set_pte_at(mm, addr, ptep, __pte(0))
Signature · 7 lines
2.25.1 _______________________________________________ linux-snps-arc mailing list [email protected] http://lists.infradead.org/mailman/listinfo/linux-snps-arc
Vineet Gupta
maintainerPATCH v2 8/19 ARC: mm: switch pgtable_t back to struct page *
In reply to Vineet GuptaSo far ARC pgtable_t has not been struct page based to avoid extra page_address() calls involved. However the differences are down to noise and get in the way of using generic code, hence this patch.
--- arch/arc/include/asm/page.h | 2 +- arch/arc/include/asm/pgalloc.h | 57 ++++++++++------------------------ arch/arc/mm/init.c | 3 ++ 3 files changed, 21 insertions(+), 41 deletions(-)
arch/arc/include/asm/page.h +1 -1
@@ -58,7 +58,7 @@ typedef struct {#define pte_pgprot (x) __pgprot(pte_val(x)) -typedef pte_t *pgtable_t ; +typedef struct page *pgtable_t ;/* * Use virt_to_pfn with caution:
arch/arc/include/asm/pgalloc.h +17 -40
100 lines not on this page · raw message
arch/arc/mm/init.c +3 -0
@@ -189,6 +189,9 @@ void __init mem_init(void) {memblock_free_all ();highmem_init (); + +BUILD_BUG_ON ((PTRS_PER_PGD *sizeof (pgd_t)) >PAGE_SIZE ); +BUILD_BUG_ON ((PTRS_PER_PTE *sizeof (pte_t)) >PAGE_SIZE ); }#ifdef CONFIG_HIGHMEM
Signature · 7 lines
2.25.1 _______________________________________________ linux-snps-arc mailing list [email protected] http://lists.infradead.org/mailman/listinfo/linux-snps-arc
kernel test robot
build botRe: [PATCH v2 08/19] ARC: mm: switch pgtable_t back to struct page *
I love your patch! Yet something to improve:
Hi Vineet,
I love your patch! Yet something to improve:
[auto build test ERROR on arc/for-next]
[also build test ERROR on linux/master linus/master v5.14-rc5 next-20210812] [If your patch is applied to the wrong git tree, kindly drop us a note. And when submitting patch, we suggest to use '--base' as documented in
https://git-scm.com/docs/git-format-patch]
url: https://github.com/0day-ci/linux/commits/Vineet-Gupta/ARC-mm-updates-support-3-4-levels-and-asm-generic-pgalloc/20210813-074023 base: https://git.kernel.org/pub/scm/linux/kernel/git/vgupta/arc.git for-next config: arc-randconfig-s032-20210812 (attached as .config) compiler: arceb-elf-gcc (GCC) 10.3.0
reproduce:
wget https://raw.githubusercontent.com/intel/lkp-tests/master/sbin/make.cross -O ~/bin/make.cross chmod +x ~/bin/make.cross
# apt-get install sparse
# sparse version: v0.6.3-348-gf0e6938b-dirty
# https://github.com/0day-ci/linux/commit/36e618e081c5a49b2aff51823c5f012045e902ef
git remote add linux-review https://github.com/0day-ci/linux
git fetch --no-tags linux-review Vineet-Gupta/ARC-mm-updates-support-3-4-levels-and-asm-generic-pgalloc/20210813-074023 git checkout 36e618e081c5a49b2aff51823c5f012045e902ef
# save the attached .config to linux build tree
COMPILER_INSTALL_PATH=$HOME/0day COMPILER=gcc-10.3.0 make.cross C=1 CF='-fdiagnostic-prefix -D__CHECK_ENDIAN__' ARCH=arc
If you fix the issue, kindly add following tag as appropriate
All errors (new ones prefixed by >>):
arch/arc/mm/init.c:35:13: warning: no previous prototype for 'arc_get_mem_sz' [-Wmissing-prototypes]
35 | long __init arc_get_mem_sz(void)
| ^~~~~~~~~~~~~~
arch/arc/mm/init.c:88:13: warning: no previous prototype for 'setup_arch_memory' [-Wmissing-prototypes]
88 | void __init setup_arch_memory(void)
| ^~~~~~~~~~~~~~~~~
In file included from <command-line>:
arch/arc/mm/init.c: In function 'mem_init':
>> include/linux/compiler_types.h:328:38: error: call to '__compiletime_assert_328' declared with attribute error: BUILD_BUG_ON failed: (PTRS_PER_PGD * sizeof(pgd_t)) > PAGE_SIZE
328 | _compiletime_assert(condition, msg, __compiletime_assert_, __COUNTER__)
| ^
include/linux/compiler_types.h:309:4: note: in definition of macro '__compiletime_assert'
309 | prefix ## suffix(); \
| ^~~~~~
include/linux/compiler_types.h:328:2: note: in expansion of macro '_compiletime_assert'
328 | _compiletime_assert(condition, msg, __compiletime_assert_, __COUNTER__)
| ^~~~~~~~~~~~~~~~~~~
include/linux/build_bug.h:39:37: note: in expansion of macro 'compiletime_assert'
39 | #define BUILD_BUG_ON_MSG(cond, msg) compiletime_assert(!(cond), msg)
| ^~~~~~~~~~~~~~~~~~
include/linux/build_bug.h:50:2: note: in expansion of macro 'BUILD_BUG_ON_MSG'
50 | BUILD_BUG_ON_MSG(condition, "BUILD_BUG_ON failed: " #condition)
| ^~~~~~~~~~~~~~~~
arch/arc/mm/init.c:193:2: note: in expansion of macro 'BUILD_BUG_ON'
193 | BUILD_BUG_ON((PTRS_PER_PGD * sizeof(pgd_t)) > PAGE_SIZE);
| ^~~~~~~~~~~~vim +/__compiletime_assert_328 +328 include/linux/compiler_types.h
eb5c2d4b45e3d2 Will Deacon 2020-07-21 314 eb5c2d4b45e3d2 Will Deacon 2020-07-21 315 #define _compiletime_assert(condition, msg, prefix, suffix) \ eb5c2d4b45e3d2 Will Deacon 2020-07-21 316 __compiletime_assert(condition, msg, prefix, suffix) eb5c2d4b45e3d2 Will Deacon 2020-07-21 317 eb5c2d4b45e3d2 Will Deacon 2020-07-21 318 /** eb5c2d4b45e3d2 Will Deacon 2020-07-21 319 * compiletime_assert - break build and emit msg if condition is false eb5c2d4b45e3d2 Will Deacon 2020-07-21 320 * @condition: a compile-time constant condition to check eb5c2d4b45e3d2 Will Deacon 2020-07-21 321 * @msg: a message to emit if condition is false eb5c2d4b45e3d2 Will Deacon 2020-07-21 322 * eb5c2d4b45e3d2 Will Deacon 2020-07-21 323 * In tradition of POSIX assert, this macro will break the build if the eb5c2d4b45e3d2 Will Deacon 2020-07-21 324 * supplied condition is *false*, emitting the supplied error message if the eb5c2d4b45e3d2 Will Deacon 2020-07-21 325 * compiler has support to do so. eb5c2d4b45e3d2 Will Deacon 2020-07-21 326 */ eb5c2d4b45e3d2 Will Deacon 2020-07-21 327 #define compiletime_assert(condition, msg) \
eb5c2d4b45e3d2 Will Deacon 2020-07-21 @328 _compiletime_assert(condition, msg, __compiletime_assert_, __COUNTER__)eb5c2d4b45e3d2 Will Deacon 2020-07-21 329
---
0-DAY CI Kernel Test Service, Intel Corporation
https://lists.01.org/hyperkitty/list/kbuild-all(a)lists.01.org
Vineet Gupta
maintainerRe: [PATCH v2 08/19] ARC: mm: switch pgtable_t back to struct page *
Quoted 49 lines · kernel test robot
Hi Vineet,
I love your patch! Yet something to improve:
[auto build test ERROR on arc/for-next]
[also build test ERROR on linux/master linus/master v5.14-rc5 next-20210812] [If your patch is applied to the wrong git tree, kindly drop us a note. And when submitting patch, we suggest to use '--base' as documented in
https://git-scm.com/docs/git-format-patch]url: https://github.com/0day-ci/linux/commits/Vineet-Gupta/ARC-mm-updates-support-3-4-levels-and-asm-generic-pgalloc/20210813-074023 base: https://git.kernel.org/pub/scm/linux/kernel/git/vgupta/arc.git for-next config: arc-randconfig-s032-20210812 (attached as .config) compiler: arceb-elf-gcc (GCC) 10.3.0reproduce:
wget https://raw.githubusercontent.com/intel/lkp-tests/master/sbin/make.cross -O ~/bin/make.cross chmod +x ~/bin/make.cross
# apt-get install sparse
# sparse version: v0.6.3-348-gf0e6938b-dirty
# https://github.com/0day-ci/linux/commit/36e618e081c5a49b2aff51823c5f012045e902ef
git remote add linux-review https://github.com/0day-ci/linux
git fetch --no-tags linux-review Vineet-Gupta/ARC-mm-updates-support-3-4-levels-and-asm-generic-pgalloc/20210813-074023 git checkout36e618e081c5a49b2aff51823c5f012045e902ef
# save the attached .config to linux build tree
COMPILER_INSTALL_PATH=$HOME/0day COMPILER=gcc-10.3.0 make.cross C=1 CF='-fdiagnostic-prefix -D__CHECK_ENDIAN__' ARCH=arcIf you fix the issue, kindly add following tag as appropriate
Reported-by kernel test robot <[email protected]>All errors (new ones prefixed by >>):
arch/arc/mm/init.c:35:13: warning: no previous prototype for 'arc_get_mem_sz' [-Wmissing-prototypes]
35 | long__initarc_get_mem_sz(void)
| ^~~~~~~~~~~~~~
arch/arc/mm/init.c:88:13: warning: no previous prototype for 'setup_arch_memory' [-Wmissing-prototypes]
88 | void__initsetup_arch_memory(void)
| ^~~~~~~~~~~~~~~~~
In file included from <command-line>:
arch/arc/mm/init.c: In function 'mem_init':Quoted 1 line
Quoted 1 line
include/linux/compiler_types.h:328:38: error: call to '__compiletime_assert_328' declared with attribute error:BUILD_BUG_ONfailed: (PTRS_PER_PGD*sizeof(pgd_t)) >PAGE_SIZE328 |
_compiletime_assert(condition, msg, __compiletime_assert_, __COUNTER__)
| ^
include/linux/compiler_types.h:309:4: note: in definition of macro '__compiletime_assert'
309 | prefix ##suffix(); \
| ^~~~~~
include/linux/compiler_types.h:328:2: note: in expansion of macro '_compiletime_assert'
328 |_compiletime_assert(condition, msg, __compiletime_assert_, __COUNTER__)
| ^~~~~~~~~~~~~~~~~~~
include/linux/build_bug.h:39:37: note: in expansion of macro 'compiletime_assert'
Quoted 9 lines
39 | #define
BUILD_BUG_ON_MSG(cond, msg)compiletime_assert(!(cond), msg)
| ^~~~~~~~~~~~~~~~~~
include/linux/build_bug.h:50:2: note: in expansion of macro 'BUILD_BUG_ON_MSG'
50 |BUILD_BUG_ON_MSG(condition, "BUILD_BUG_ON failed: " #condition)
| ^~~~~~~~~~~~~~~~
arch/arc/mm/init.c:193:2: note: in expansion of macro 'BUILD_BUG_ON'
193 |BUILD_BUG_ON((PTRS_PER_PGD * sizeof(pgd_t)) > PAGE_SIZE);
| ^~~~~~~~~~~~
Thx for the report. This is 4K page config, where PGDIR_SHIFT 21 causes 32-21 = 11 bits or 2k entries which won't fit in 4k page. I've added the fix and added Tested-by.
-Vineet
Vineet Gupta
maintainerPATCH v2 9/19 ARC: mm: switch to asm-generic/pgalloc.h
In reply to Vineet GuptaWith previous patch ARC pgalloc functions are same as generic, hence switch to that.
--- arch/arc/include/asm/pgalloc.h | 42 +--------------------------------- 1 file changed, 1 insertion(+), 41 deletions(-)
arch/arc/include/asm/pgalloc.h +1 -41
@@ -31,6 +31,7 @@#include <linux/mm.h> #include <linux/log2.h> +#include <asm-generic/pgalloc.h> static inline void pmd_populate_kernel (struct mm_struct *mm,pmd_t *pmd,pte_t *pte)
@@ -69,47 +70,6 @@ static inline pgd_t *pgd_alloc(struct mm_struct *mm) return ret; } -static inline void pgd_free (struct mm_struct *mm,pgd_t *pgd) -{ -free_page ((unsigned long )pgd); -} - -static inline pte_t *pte_alloc_one_kernel (struct mm_struct *mm) -{ -pte_t *pte; - - pte = (pte_t *)__get_free_page (GFP_KERNEL | __GFP_ZERO); - -return pte; -} - -static inline pgtable_t pte_alloc_one (struct mm_struct *mm) -{ -struct page *page; - - page = (pgtable_t )alloc_page (GFP_KERNEL | __GFP_ZERO | __GFP_ACCOUNT); -if (!page) -return NULL ; - -if (!pgtable_pte_page_ctor (page)) { -__free_page (page); -return NULL ; - } - -return page; -} - -static inline void pte_free_kernel (struct mm_struct *mm,pte_t *pte) -{ -free_page ((unsigned long )pte); -} - -static inline void pte_free (struct mm_struct *mm,pgtable_t pte_page) -{ -pgtable_pte_page_dtor (pte_page); -__free_page (pte_page); -} -#define __pte_free_tlb (tlb, pte, addr) pte_free((tlb)->mm, pte)#endif /* _ASM_ARC_PGALLOC_H */
Signature · 7 lines
2.25.1 _______________________________________________ linux-snps-arc mailing list [email protected] http://lists.infradead.org/mailman/listinfo/linux-snps-arc
Vineet Gupta
maintainerPATCH v2 10/19 ARC: mm: non-functional code cleanup ahead of 3 levels
In reply to Vineet Gupta--- arch/arc/include/asm/page.h | 30 ++++++++++++++++-------------- 1 file changed, 16 insertions(+), 14 deletions(-)
arch/arc/include/asm/page.h +16 -14
@@ -34,6 +34,13 @@ void copy_user_highpage(struct page *to, struct page *from, unsigned long u_vaddr, struct vm_area_struct *vma);void clear_user_page (void *to,unsigned long u_vaddr,struct page *page); +typedef struct { +unsigned long pgd; +}pgd_t ; + +#define pgd_val (x) ((x).pgd) +#define __pgd (x) ((pgd_t) { (x) }) + typedef struct { #ifdef CONFIG_ARC_HAS_PAE40 unsigned long long pte;
@@ -41,22 +48,17 @@ typedef struct {unsigned long pte;#endif } pte_t; -typedef struct { -unsigned long pgd; -}pgd_t ; + +#define pte_val (x) ((x).pte) +#define __pte (x) ((pte_t) { (x) }) +typedef struct {unsigned long pgprot; }pgprot_t ; -#define pte_val (x) ((x).pte) -#define pgd_val (x) ((x).pgd) -#define pgprot_val (x) ((x).pgprot) - -#define __pte (x) ((pte_t) { (x) }) -#define __pgd (x) ((pgd_t) { (x) }) -#define __pgprot (x) ((pgprot_t) { (x) }) - -#define pte_pgprot (x) __pgprot(pte_val(x)) +#define pgprot_val (x) ((x).pgprot) +#define __pgprot (x) ((pgprot_t) { (x) }) +#define pte_pgprot (x) __pgprot(pte_val(x))typedef struct page *pgtable_t ;
@@ -96,8 +98,8 @@ extern int pfn_valid(unsigned long pfn);* virt here means link-address/program-address as embedded in object code. * And for ARC, link-addr = physical address */ -#define __pa (vaddr) ((unsigned long)(vaddr)) -#define __va (paddr) ((void *)((unsigned long)(paddr))) +#define __pa (vaddr) ((unsigned long)(vaddr)) +#define __va (paddr) ((void *)((unsigned long)(paddr)))#define virt_to_page (kaddr) pfn_to_page(virt_to_pfn(kaddr))#define virt_addr_valid (kaddr) pfn_valid(virt_to_pfn(kaddr))
Signature · 7 lines
2.25.1 _______________________________________________ linux-snps-arc mailing list [email protected] http://lists.infradead.org/mailman/listinfo/linux-snps-arc
Vineet Gupta
maintainerPATCH v2 11/19 ARC: mm: move MMU specific bits out of ASID allocator
In reply to Vineet GuptaAnd while at it, rewrite commentary on ASID allocator
--- arch/arc/include/asm/mmu.h | 13 +++++++++++++ arch/arc/include/asm/mmu_context.h | 28 +++++++++++++--------------- arch/arc/mm/tlb.c | 11 ++++------- 3 files changed, 30 insertions(+), 22 deletions(-)
arch/arc/include/asm/mmu.h +13 -0
@@ -64,6 +64,19 @@ typedef struct {unsigned long asid[NR_CPUS ];/* 8 bit MMU PID + Generation cycle */ } mm_context_t; +static void inline mmu_setup_asid (struct mm_struct *mm,unsigned int asid) +{ +write_aux_reg (ARC_REG_PID , asid |MMU_ENABLE ); +} + +static void inline mmu_setup_pgd (struct mm_struct *mm,void *pgd) +{ +/* PGD cached in MMU reg to avoid 3 mem lookups: task->mm->pgd */ +#ifdef CONFIG_ISA_ARCV2 +write_aux_reg (ARC_REG_SCRATCH_DATA0 , (unsigned int )pgd); +#endif +} +static inline int is_pae40_enabled (void ) {return IS_ENABLED (CONFIG_ARC_HAS_PAE40 );
arch/arc/include/asm/mmu_context.h +13 -15
@@ -15,22 +15,23 @@#ifndef _ASM_ARC_MMU_CONTEXT_H#define _ASM_ARC_MMU_CONTEXT_H -#include <asm/arcregs.h> -#include <asm/tlb.h> #include <linux/sched/mm.h> +#include <asm/tlb.h> #include <asm-generic/mm_hooks.h> -/* ARC700 ASID Management +/* ARC ASID Management +* +* MMU tags TLBs with an 8-bit ASID, avoiding need to flush the TLB on +* context-switch. * -* ARC MMU provides 8-bit ASID (0..255) to TAG TLB entries, allowing entries -* with same vaddr (different tasks) to co-exit. This provides for -* "Fast Context Switch" i.e. no TLB flush on ctxt-switch +* ASID is managed per cpu, so task threads across CPUs can have different +* ASID. Global ASID management is needed if hardware supports TLB shootdown +* and/or shared TLB across cores, which ARC doesn't. * -* Linux assigns each task a unique ASID. A simple round-robin allocation -* of H/w ASID is done using software tracker @asid_cpu. -* When it reaches max 255, the allocation cycle starts afresh by flushing -* the entire TLB and wrapping ASID back to zero. +* Each task is assigned unique ASID, with a simple round-robin allocator +* tracked in @asid_cpu. When 8-bit value rolls over,a new cycle is started +* over from 0, and TLB is flushed * * A new allocation cycle, post rollover, could potentially reassign an ASID * to a different task. Thus the rule is to refresh the ASID in a new cycle.
@@ -93,7 +94,7 @@ static inline void get_new_mmu_context(struct mm_struct *mm)asid_mm (mm, cpu) =asid_cpu (cpu); set_hw: -write_aux_reg (ARC_REG_PID ,hw_pid (mm, cpu) |MMU_ENABLE ); +mmu_setup_asid (mm,hw_pid (mm, cpu));local_irq_restore (flags); }
@@ -146,10 +147,7 @@ static inline void switch_mm(struct mm_struct *prev, struct mm_struct *next,*/ cpumask_set_cpu (cpu,mm_cpumask (next)); -#ifdef CONFIG_ISA_ARCV2 -/* PGD cached in MMU reg to avoid 3 mem lookups: task->mm->pgd */ -write_aux_reg (ARC_REG_SCRATCH_DATA0 , next->pgd); -#endif +mmu_setup_pgd (next, next->pgd);get_new_mmu_context (next); }
arch/arc/mm/tlb.c +4 -7
@@ -716,14 +716,11 @@ void arc_mmu_init(void)if (IS_ENABLED (CONFIG_ARC_HAS_PAE40 ) && !mmu->pae)panic ("Hardware doesn't support PAE40\n" ); -/* Enable the MMU */ -write_aux_reg (ARC_REG_PID ,MMU_ENABLE ); +/* Enable the MMU with ASID 0 */ +mmu_setup_asid (NULL ,0 ); -/* In arc700/smp needed for re-entrant interrupt handling */ -#ifdef CONFIG_ISA_ARCV2 -/* swapper_pg_dir is the pgd for the kernel, used by vmalloc */ -write_aux_reg (ARC_REG_SCRATCH_DATA0 , swapper_pg_dir); -#endif +/* cache the pgd pointer in MMU SCRATCH reg (ARCv2 only) */ +mmu_setup_pgd (NULL , swapper_pg_dir);if (pae40_exist_but_not_enab ())write_aux_reg (ARC_REG_TLBPD1HI ,0 );
Signature · 7 lines
2.25.1 _______________________________________________ linux-snps-arc mailing list [email protected] http://lists.infradead.org/mailman/listinfo/linux-snps-arc
Vineet Gupta
maintainerPATCH v2 12/19 ARC: mm: move MMU specific bits out of entry code ...
In reply to Vineet Gupta... to avoid polluting shared entry code (across three ISA variants) with ISA/MMU specific code.
--- arch/arc/include/asm/mmu.h | 8 ++++++++ arch/arc/kernel/entry-arcv2.S | 1 + arch/arc/kernel/entry.S | 7 ++----- 3 files changed, 11 insertions(+), 5 deletions(-)
arch/arc/include/asm/mmu.h +8 -0
@@ -84,6 +84,14 @@ static inline int is_pae40_enabled(void) extern int pae40_exist_but_not_enab(void); +#else + +.macro ARC_MMU_REENABLE reg + lr \reg, [ARC_REG_PID] + or \reg, \reg, MMU_ENABLE + sr \reg, [ARC_REG_PID] +.endm + #endif /* !__ASSEMBLY__ */ #endif
arch/arc/kernel/entry-arcv2.S +1 -0
@@ -10,6 +10,7 @@ #include <asm/errno.h> #include <asm/arcregs.h> #include <asm/irqflags.h> +#include <asm/mmu.h> ; A maximum number of supported interrupts in the core interrupt controller. ; This number is not equal to the maximum interrupt number (256) because
arch/arc/kernel/entry.S +2 -5
@@ -101,11 +101,8 @@ ENTRY(EV_MachineCheck) lr r0, [efa] mov r1, sp - ; hardware auto-disables MMU, re-enable it to allow kernel vaddr - ; access for say stack unwinding of modules for crash dumps - lr r3, [ARC_REG_PID] - or r3, r3, MMU_ENABLE - sr r3, [ARC_REG_PID] + ; MC excpetions disable MMU + ARC_MMU_REENABLE r3 lsr r3, r2, 8 bmsk r3, r3, 7
Signature · 7 lines
2.25.1 _______________________________________________ linux-snps-arc mailing list [email protected] http://lists.infradead.org/mailman/listinfo/linux-snps-arc
Vineet Gupta
maintainerPATCH v2 13/19 ARC: mm: disintegrate mmu.h (arcv2 bits out)
In reply to Vineet Guptanon functional change
--- arch/arc/include/asm/mmu-arcv2.h | 103 +++++++++++++++++++++++++++++++ arch/arc/include/asm/mmu.h | 80 +----------------------- arch/arc/include/asm/pgtable.h | 6 -- 3 files changed, 105 insertions(+), 84 deletions(-) create mode 100644 arch/arc/include/asm/mmu-arcv2.h
arch/arc/include/asm/mmu-arcv2.h new file +103 -0
@@ -0,0 +1,103 @@ +/* SPDX-License-Identifier: GPL-2.0-only */ +/* +* Copyright (C) 2004, 2007-2010, 2011-2012, 2019-20 Synopsys, Inc. (www.synopsys.com) +* +* MMUv3 (arc700) / MMUv4 (archs) are software page walked and software managed. +* This file contains the TLB access registers and commands +*/ + +#ifndef _ASM_ARC_MMU_ARCV2_H +#define _ASM_ARC_MMU_ARCV2_H + +/* + * TLB Management regs + */ +#define ARC_REG_MMU_BCR 0x06f + +#ifdef CONFIG_ARC_MMU_V3 +#define ARC_REG_TLBPD0 0x405 +#define ARC_REG_TLBPD1 0x406 +#define ARC_REG_TLBPD1HI 0 /* Dummy: allows common code */ +#define ARC_REG_TLBINDEX 0x407 +#define ARC_REG_TLBCOMMAND 0x408 +#define ARC_REG_PID 0x409 +#define ARC_REG_SCRATCH_DATA0 0x418 +#else +#define ARC_REG_TLBPD0 0x460 +#define ARC_REG_TLBPD1 0x461 +#define ARC_REG_TLBPD1HI 0x463 +#define ARC_REG_TLBINDEX 0x464 +#define ARC_REG_TLBCOMMAND 0x465 +#define ARC_REG_PID 0x468 +#define ARC_REG_SCRATCH_DATA0 0x46c +#endif + +/* Bits in MMU PID reg */ +#define __TLB_ENABLE (1 << 31) +#define __PROG_ENABLE (1 << 30) +#define MMU_ENABLE (__TLB_ENABLE | __PROG_ENABLE) + +/* Bits in TLB Index reg */ +#define TLB_LKUP_ERR 0x80000000 + +#ifdef CONFIG_ARC_MMU_V3 +#define TLB_DUP_ERR (TLB_LKUP_ERR | 0x00000001) +#else +#define TLB_DUP_ERR (TLB_LKUP_ERR | 0x40000000) +#endif + +/* + * TLB Commands + */ +#define TLBWrite 0x1 +#define TLBRead 0x2 +#define TLBGetIndex 0x3 +#define TLBProbe 0x4 +#define TLBWriteNI 0x5 /* write JTLB without inv uTLBs */ +#define TLBIVUTLB 0x6 /* explicitly inv uTLBs */ + +#ifdef CONFIG_ARC_MMU_V4 +#define TLBInsertEntry 0x7 +#define TLBDeleteEntry 0x8 +#endif + +/* Masks for actual TLB "PD"s */ +#define PTE_BITS_IN_PD0 (_PAGE_GLOBAL | _PAGE_PRESENT | _PAGE_HW_SZ) +#define PTE_BITS_RWX (_PAGE_EXECUTE | _PAGE_WRITE | _PAGE_READ) + +#define PTE_BITS_NON_RWX_IN_PD1 (PAGE_MASK_PHYS | _PAGE_CACHEABLE) + +#ifndef __ASSEMBLY__ + +struct mm_struct; +extern int pae40_exist_but_not_enab(void); + +static inline int is_pae40_enabled(void) +{ + return IS_ENABLED(CONFIG_ARC_HAS_PAE40); +} + +static void inline mmu_setup_asid(struct mm_struct *mm, unsigned long asid) +{ + write_aux_reg(ARC_REG_PID, asid | MMU_ENABLE); +} + +static void inline mmu_setup_pgd(struct mm_struct *mm, void *pgd) +{ + /* PGD cached in MMU reg to avoid 3 mem lookups: task->mm->pgd */ +#ifdef CONFIG_ISA_ARCV2 + write_aux_reg(ARC_REG_SCRATCH_DATA0, (unsigned int)pgd); +#endif +} + +#else + +.macro ARC_MMU_REENABLE reg + lr \reg, [ARC_REG_PID] + or \reg, \reg, MMU_ENABLE + sr \reg, [ARC_REG_PID] +.endm + +#endif /* !__ASSEMBLY__ */ + +#endif
arch/arc/include/asm/mmu.h +2 -78
93 lines not on this page · raw message
arch/arc/include/asm/pgtable.h +0 -6
@@ -103,12 +103,6 @@*/ #define PAGE_KERNEL __pgprot(_K_PAGE_PERMS | _PAGE_CACHEABLE) -/* Masks for actual TLB "PD"s */ -#define PTE_BITS_IN_PD0 (_PAGE_GLOBAL | _PAGE_PRESENT | _PAGE_HW_SZ) -#define PTE_BITS_RWX (_PAGE_EXECUTE | _PAGE_WRITE | _PAGE_READ) - -#define PTE_BITS_NON_RWX_IN_PD1 (PAGE_MASK_PHYS | _PAGE_CACHEABLE) -/************************************************************************** * Mapping of vm_flags (Generic VM) to PTE flags (arch specific) *
Signature · 7 lines
2.25.1 _______________________________________________ linux-snps-arc mailing list [email protected] http://lists.infradead.org/mailman/listinfo/linux-snps-arc
kernel test robot
build botRe: [PATCH v2 13/19] ARC: mm: disintegrate mmu.h (arcv2 bits out)
I love your patch! Yet something to improve:
Hi Vineet,
I love your patch! Yet something to improve:
[auto build test ERROR on arc/for-next]
[also build test ERROR on next-20210812]
[cannot apply to linux/master linus/master v5.14-rc5]
[If your patch is applied to the wrong git tree, kindly drop us a note. And when submitting patch, we suggest to use '--base' as documented in
https://git-scm.com/docs/git-format-patch]
url: https://github.com/0day-ci/linux/commits/Vineet-Gupta/ARC-mm-updates-support-3-4-levels-and-asm-generic-pgalloc/20210813-074023 base: https://git.kernel.org/pub/scm/linux/kernel/git/vgupta/arc.git for-next config: arc-allyesconfig (attached as .config) compiler: arceb-elf-gcc (GCC) 10.3.0
reproduce (this is a W=1 build):
wget https://raw.githubusercontent.com/intel/lkp-tests/master/sbin/make.cross -O ~/bin/make.cross chmod +x ~/bin/make.cross
# https://github.com/0day-ci/linux/commit/0411d3a95cb73722d026f7b3d9c9d8abab8c0d79
git remote add linux-review https://github.com/0day-ci/linux
git fetch --no-tags linux-review Vineet-Gupta/ARC-mm-updates-support-3-4-levels-and-asm-generic-pgalloc/20210813-074023 git checkout 0411d3a95cb73722d026f7b3d9c9d8abab8c0d79
# save the attached .config to linux build tree
COMPILER_INSTALL_PATH=$HOME/0day COMPILER=gcc-10.3.0 make.cross ARCH=arc
If you fix the issue, kindly add following tag as appropriate
All errors (new ones prefixed by >>):
In file included from arch/arc/include/asm/mmu.h:19,
from include/linux/mm_types.h:19, from include/linux/buildid.h:5, from include/linux/module.h:14, from lib/test_bitops.c:9:
>> arch/arc/include/asm/mmu-arcv2.h:80:1: error: 'inline' is not at beginning of declaration [-Werror=old-style-declaration]
80 | static void inline mmu_setup_asid(struct mm_struct *mm, unsigned long asid)
| ^~~~~~
arch/arc/include/asm/mmu-arcv2.h:85:1: error: 'inline' is not at beginning of declaration [-Werror=old-style-declaration]
85 | static void inline mmu_setup_pgd(struct mm_struct *mm, void *pgd)
| ^~~~~~
cc1: all warnings being treated as errors
vim +/inline +80 arch/arc/include/asm/mmu-arcv2.h
79
> 80 static void inline mmu_setup_asid(struct mm_struct *mm, unsigned long asid)
81 {
82 write_aux_reg(ARC_REG_PID, asid | MMU_ENABLE);
83 }
84 ---
0-DAY CI Kernel Test Service, Intel Corporation
https://lists.01.org/hyperkitty/list/kbuild-all(a)lists.01.org
Vineet Gupta
maintainerRe: [PATCH v2 13/19] ARC: mm: disintegrate mmu.h (arcv2 bits out)
Quoted 35 lines · kernel test robot
Hi Vineet,
I love your patch! Yet something to improve:
[auto build test ERROR on arc/for-next]
[also build test ERROR on next-20210812]
[cannot apply to linux/master linus/master v5.14-rc5]
[If your patch is applied to the wrong git tree, kindly drop us a note. And when submitting patch, we suggest to use '--base' as documented in
https://git-scm.com/docs/git-format-patch]url:https://github.com/0day-ci/linux/commits/Vineet-Gupta/ARC-mm-updates-support-3-4-levels-and-asm-generic-pgalloc/20210813-074023 base:https://git.kernel.org/pub/scm/linux/kernel/git/vgupta/arc.git for-next
config: arc-allyesconfig (attached as .config)
compiler: arceb-elf-gcc (GCC) 10.3.0
reproduce (this is a W=1 build):
wgethttps://raw.githubusercontent.com/intel/lkp-tests/master/sbin/make.cross -O ~/bin/make.cross chmod +x ~/bin/make.cross #https://github.com/0day-ci/linux/commit/0411d3a95cb73722d026f7b3d9c9d8abab8c0d79
git remote add linux-reviewhttps://github.com/0day-ci/linux
git fetch --no-tags linux-review Vineet-Gupta/ARC-mm-updates-support-3-4-levels-and-asm-generic-pgalloc/20210813-074023 git checkout0411d3a95cb73722d026f7b3d9c9d8abab8c0d79
# save the attached .config to linux build tree
COMPILER_INSTALL_PATH=$HOME/0day COMPILER=gcc-10.3.0 make.cross ARCH=arcIf you fix the issue, kindly add following tag as appropriate
Reported-by kernel test robot<[email protected]>All errors (new ones prefixed by >>):
In file included from
arch/arc/include/asm/mmu.h:19,
frominclude/linux/mm_types.h:19, frominclude/linux/buildid.h:5, frominclude/linux/module.h:14, fromlib/test_bitops.c:9:
Quoted 7 lines
Quoted 1 line
Quoted 1 line
arch/arc/include/asm/mmu-arcv2.h:80:1: error: 'inline' is not at beginning of declaration [-Werror=old-style-declaration]80 | static void inline
mmu_setup_asid(struct mm_struct *mm, unsigned long asid)
| ^~~~~~
arch/arc/include/asm/mmu-arcv2.h:85:1: error: 'inline' is not at beginning of declaration [-Werror=old-style-declaration]
85 | static void inlinemmu_setup_pgd(struct mm_struct *mm, void *pgd)
| ^~~~~~
cc1: all warnings being treated as errors
Strange my gcc11 is not tripping on this snafu. Fixed now. Thx for the report as always.
-Vineet
Vineet Gupta
maintainerPATCH v2 14/19 ARC: mm: disintegrate pgtable.h into levels and flags
In reply to Vineet Gupta - pgtable-bits-arcv2.h (MMU specific page table flags)
- pgtable-levels.h (paging levels)
No functional changes, but paves way for easy addition of new MMU code with different bits and levels etc
--- arch/arc/include/asm/pgtable-bits-arcv2.h | 149 ++++++++++++ arch/arc/include/asm/pgtable-levels.h | 91 +++++++ arch/arc/include/asm/pgtable.h | 277 +--------------------- 3 files changed, 244 insertions(+), 273 deletions(-) create mode 100644 arch/arc/include/asm/pgtable-bits-arcv2.h create mode 100644 arch/arc/include/asm/pgtable-levels.h
arch/arc/include/asm/pgtable-bits-arcv2.h new file +149 -0
@@ -0,0 +1,149 @@ +/* SPDX-License-Identifier: GPL-2.0-only */ +/* +* Copyright (C) 2004, 2007-2010, 2011-2012 Synopsys, Inc. (www.synopsys.com) +*/ + +/* +* page table flags for software walked/managed MMUv3 (ARC700) and MMUv4 (HS) +* There correspond to the corresponding bits in the TLB +*/ + +#ifndef _ASM_ARC_PGTABLE_BITS_ARCV2_H +#define _ASM_ARC_PGTABLE_BITS_ARCV2_H + +#ifdef CONFIG_ARC_CACHE_PAGES +#define _PAGE_CACHEABLE (1 << 0)/* Cached (H) */ +#else +#define _PAGE_CACHEABLE 0 +#endif + +#define _PAGE_EXECUTE (1 << 1)/* User Execute (H) */ +#define _PAGE_WRITE (1 << 2)/* User Write (H) */ +#define _PAGE_READ (1 << 3)/* User Read (H) */ +#define _PAGE_ACCESSED (1 << 4)/* Accessed (s) */ +#define _PAGE_DIRTY (1 << 5)/* Modified (s) */ +#define _PAGE_SPECIAL (1 << 6) +#define _PAGE_GLOBAL (1 << 8)/* ASID agnostic (H) */ +#define _PAGE_PRESENT (1 << 9)/* PTE/TLB Valid (H) */ + +#ifdef CONFIG_ARC_MMU_V4 +#define _PAGE_HW_SZ (1 << 10)/* Normal/super (H) */ +#else +#define _PAGE_HW_SZ 0 +#endif + +/* Defaults for every user page */ +#define ___DEF (_PAGE_PRESENT | _PAGE_CACHEABLE) + +/* Set of bits not changed in pte_modify */ +#define _PAGE_CHG_MASK (PAGE_MASK_PHYS | _PAGE_ACCESSED | _PAGE_DIRTY | \ + _PAGE_SPECIAL) + +/* More Abbrevaited helpers */ +#define PAGE_U_NONE __pgprot(___DEF) +#define PAGE_U_R __pgprot(___DEF | _PAGE_READ) +#define PAGE_U_W_R __pgprot(___DEF | _PAGE_READ | _PAGE_WRITE) +#define PAGE_U_X_R __pgprot(___DEF | _PAGE_READ | _PAGE_EXECUTE) +#define PAGE_U_X_W_R __pgprot(___DEF \ + | _PAGE_READ | _PAGE_WRITE | _PAGE_EXECUTE) +#define PAGE_KERNEL __pgprot(___DEF | _PAGE_GLOBAL \ + | _PAGE_READ | _PAGE_WRITE | _PAGE_EXECUTE) + +#define PAGE_SHARED PAGE_U_W_R + +#define pgprot_noncached (prot) (__pgprot(pgprot_val(prot) & ~_PAGE_CACHEABLE)) + +/* +* Mapping of vm_flags (Generic VM) to PTE flags (arch specific) +* +* Certain cases have 1:1 mapping +* e.g. __P101 means VM_READ, VM_EXEC and !VM_SHARED +* which directly corresponds to PAGE_U_X_R +* +* Other rules which cause the divergence from 1:1 mapping +* +* 1. Although ARC700 can do exclusive execute/write protection (meaning R +* can be tracked independet of X/W unlike some other CPUs), still to +* keep things consistent with other archs: +* -Write implies Read: W => R +* -Execute implies Read: X => R +* +* 2. Pvt Writable doesn't have Write Enabled initially: Pvt-W => !W +* This is to enable COW mechanism +*/ +/* xwr */ +#define __P000 PAGE_U_NONE +#define __P001 PAGE_U_R +#define __P010 PAGE_U_R/* Pvt-W => !W */ +#define __P011 PAGE_U_R/* Pvt-W => !W */ +#define __P100 PAGE_U_X_R/* X => R */ +#define __P101 PAGE_U_X_R +#define __P110 PAGE_U_X_R/* Pvt-W => !W and X => R */ +#define __P111 PAGE_U_X_R/* Pvt-W => !W */ + +#define __S000 PAGE_U_NONE +#define __S001 PAGE_U_R +#define __S010 PAGE_U_W_R/* W => R */ +#define __S011 PAGE_U_W_R +#define __S100 PAGE_U_X_R/* X => R */ +#define __S101 PAGE_U_X_R +#define __S110 PAGE_U_X_W_R/* X => R */ +#define __S111 PAGE_U_X_W_R + +#ifndef __ASSEMBLY__ + +#define pte_write (pte) (pte_val(pte) & _PAGE_WRITE) +#define pte_dirty (pte) (pte_val(pte) & _PAGE_DIRTY) +#define pte_young (pte) (pte_val(pte) & _PAGE_ACCESSED) +#define pte_special (pte) (pte_val(pte) & _PAGE_SPECIAL) + +#define PTE_BIT_FUNC (fn, op) \ + static inline pte_t pte_##fn(pte_t pte) { pte_val(pte) op; return pte; } + +PTE_BIT_FUNC(mknotpresent, &= ~(_PAGE_PRESENT)); +PTE_BIT_FUNC(wrprotect, &= ~(_PAGE_WRITE)); +PTE_BIT_FUNC(mkwrite, |= (_PAGE_WRITE)); +PTE_BIT_FUNC(mkclean, &= ~(_PAGE_DIRTY)); +PTE_BIT_FUNC(mkdirty, |= (_PAGE_DIRTY)); +PTE_BIT_FUNC(mkold, &= ~(_PAGE_ACCESSED)); +PTE_BIT_FUNC(mkyoung, |= (_PAGE_ACCESSED)); +PTE_BIT_FUNC(mkspecial, |= (_PAGE_SPECIAL)); +PTE_BIT_FUNC(mkhuge, |= (_PAGE_HW_SZ)); + +static inline pte_t pte_modify (pte_t pte,pgprot_t newprot) +{ +return __pte ((pte_val (pte) & _PAGE_CHG_MASK) |pgprot_val (newprot)); +} + +static inline void set_pte_at (struct mm_struct *mm,unsigned long addr, +pte_t *ptep,pte_t pteval) +{ +set_pte (ptep, pteval); +} + +void update_mmu_cache (struct vm_area_struct *vma,unsigned long address, +pte_t *ptep); + +/* Encode swap {type,off} tuple into PTE +* We reserve 13 bits for 5-bit @type, keeping bits 12-5 zero, ensuring that +* PAGE_PRESENT is zero in a PTE holding swap "identifier" +*/ +#define __swp_entry (type, off) ((swp_entry_t) \ + { ((type) & 0x1f) | ((off) << 13) }) + +/* Decode a PTE containing swap "identifier "into constituents */ +#define __swp_type (pte_lookalike) (((pte_lookalike).val) & 0x1f) +#define __swp_offset (pte_lookalike) ((pte_lookalike).val >> 13) + +#define __pte_to_swp_entry (pte) ((swp_entry_t) { pte_val(pte) }) +#define __swp_entry_to_pte (x) ((pte_t) { (x).val }) + +#define kern_addr_valid (addr) (1) + +#ifdef CONFIG_TRANSPARENT_HUGEPAGE +#include <asm/hugepage.h> +#endif + +#endif /* __ASSEMBLY__ */ + +#endif
arch/arc/include/asm/pgtable-levels.h new file +91 -0
91 lines not on this page · raw message
arch/arc/include/asm/pgtable.h +4 -273
308 lines not on this page · raw message
Signature · 7 lines
2.25.1 _______________________________________________ linux-snps-arc mailing list [email protected] http://lists.infradead.org/mailman/listinfo/linux-snps-arc
Vineet Gupta
maintainerPATCH v2 15/19 ARC: mm: hack to allow 2 level build with 4 level code
In reply to Vineet GuptaPMD_SHIFT is mapped to PUD_SHIFT or PGD_SHIFT by asm-generic/pgtable-* but only for !__ASSEMBLY__
tlbex.S asm code has PTRS_PER_PTE which uses PMD_SHIFT hence barfs for CONFIG_PGTABLE_LEVEL={2,3} and works for 4.
So add a workaround local to tlbex.S - the proper fix is to change asm-generic/pgtable-* headers to expose the defines for __ASSEMBLY__ too
--- arch/arc/mm/tlbex.S | 8 ++++++++ 1 file changed, 8 insertions(+)
arch/arc/mm/tlbex.S +8 -0
@@ -145,6 +145,14 @@ ex_saved_reg1: ;TLB Miss handling Code ;============================================================================ +#ifndef PMD_SHIFT +#define PMD_SHIFT PUD_SHIFT +#endif + +#ifndef PUD_SHIFT +#define PUD_SHIFT PGDIR_SHIFT +#endif + ;----------------------------------------------------------------------------- ; This macro does the page-table lookup for the faulting address. ; OUT: r0 = PTE faulted on, r1 = ptr to PTE, r2 = Faulting V-address
Signature · 7 lines
2.25.1 _______________________________________________ linux-snps-arc mailing list [email protected] http://lists.infradead.org/mailman/listinfo/linux-snps-arc
Vineet Gupta
maintainerPATCH v2 16/19 ARC: mm: support 3 levels of page tables
In reply to Vineet GuptaARCv2 MMU is software walked and Linux implements 2 levels of paging: pgd/pte. Forthcoming hw will have multiple levels, so this change preps mm code for same. It is also fun to try multi levels even on soft-walked code to ensure generic mm code is robust to handle.
overview
________
2 levels {pgd, pte} : pmd is folded but pmd_* macros are valid and operate on pgd 3 levels {pgd, pmd, pte}:
- pud is folded and pud_* macros point to pgd
- pmd_* macros operate on actual pmd
code changes
____________
1. #include <asm-generic/pgtable-nopud.h>
2. Define CONFIG_PGTABLE_LEVELS 3
3a. Define PMD_SHIFT, PMD_SIZE, PMD_MASK, pmd_t 3b. Define pmd_val() which actually deals with pmd
(pmd_offset(), pmd_index() are provided by generic code)
3c. pmd_alloc_one()/pmd_free() also provided by generic code
(pmd_populate/pmd_free already exist)
4. Define pud_none(), pud_bad() macros based on generic pud_val() which internally pertains to pgd now.
4b. define pud_populate() to just setup pgd
--- arch/arc/Kconfig | 4 ++ arch/arc/include/asm/page.h | 11 +++++ arch/arc/include/asm/pgalloc.h | 11 +++++ arch/arc/include/asm/pgtable-levels.h | 63 ++++++++++++++++++++++++--- arch/arc/include/asm/processor.h | 2 +- arch/arc/mm/fault.c | 4 ++ arch/arc/mm/init.c | 1 + arch/arc/mm/tlb.c | 4 +- arch/arc/mm/tlbex.S | 9 ++++ 9 files changed, 101 insertions(+), 8 deletions(-)
arch/arc/Kconfig +4 -0
@@ -314,6 +314,10 @@ config ARC_HUGEPAGE_16M endchoice +config PGTABLE_LEVELS + int "Number of Page table levels" + default 2 + config ARC_COMPACT_IRQ_LEVELS depends on ISA_ARCOMPACT bool "Setup Timer IRQ as high Priority"
arch/arc/include/asm/page.h +11 -0
@@ -41,6 +41,17 @@ typedef struct {#define pgd_val (x) ((x).pgd)#define __pgd (x) ((pgd_t) { (x) }) +#if CONFIG_PGTABLE_LEVELS >2 + +typedef struct { +unsigned long pmd; +}pmd_t ; + +#define pmd_val (x) ((x).pmd) +#define __pmd (x) ((pmd_t) { (x) }) + +#endif + typedef struct { #ifdef CONFIG_ARC_HAS_PAE40 unsigned long long pte;
arch/arc/include/asm/pgalloc.h +11 -0
@@ -70,6 +70,17 @@ static inline pgd_t *pgd_alloc(struct mm_struct *mm) return ret; } +#if CONFIG_PGTABLE_LEVELS >2 + +static inline void pud_populate (struct mm_struct *mm,pud_t *pudp,pmd_t *pmdp) +{ +set_pud (pudp,__pud ((unsigned long )pmdp)); +} + +#define __pmd_free_tlb (tlb, pmd, addr) pmd_free((tlb)->mm, pmd) + +#endif +#define __pte_free_tlb (tlb, pte, addr) pte_free((tlb)->mm, pte)#endif /* _ASM_ARC_PGALLOC_H */
arch/arc/include/asm/pgtable-levels.h +58 -5
95 lines not on this page · raw message
arch/arc/include/asm/processor.h +1 -1
@@ -93,7 +93,7 @@ extern unsigned int get_wchan(struct task_struct *p);#define VMALLOC_START (PAGE_OFFSET - (CONFIG_ARC_KVADDR_SIZE << 20))/* 1 PGDIR_SIZE each for fixmap/pkmap, 2 PGDIR_SIZE gutter (see asm/highmem.h) */ -#define VMALLOC_SIZE ((CONFIG_ARC_KVADDR_SIZE << 20) - PGDIR_SIZE * 4) +#define VMALLOC_SIZE ((CONFIG_ARC_KVADDR_SIZE << 20) - PMD_SIZE * 4)#define VMALLOC_END (VMALLOC_START + VMALLOC_SIZE)
arch/arc/mm/fault.c +4 -0
@@ -39,6 +39,8 @@ noinline static int handle_kernel_vaddr_fault(unsigned long address)if (!pgd_present (*pgd_k)) goto bad_area; +set_pgd (pgd, *pgd_k); + p4d =p4d_offset (pgd, address); p4d_k =p4d_offset (pgd_k, address);if (!p4d_present (*p4d_k))
@@ -49,6 +51,8 @@ noinline static int handle_kernel_vaddr_fault(unsigned long address)if (!pud_present (*pud_k)) goto bad_area; +set_pud (pud, *pud_k); + pmd =pmd_offset (pud, address); pmd_k =pmd_offset (pud_k, address);if (!pmd_present (*pmd_k))
arch/arc/mm/init.c +1 -0
@@ -191,6 +191,7 @@ void __init mem_init(void)highmem_init ();BUILD_BUG_ON ((PTRS_PER_PGD *sizeof (pgd_t)) >PAGE_SIZE ); +BUILD_BUG_ON ((PTRS_PER_PMD *sizeof (pmd_t)) >PAGE_SIZE );BUILD_BUG_ON ((PTRS_PER_PTE *sizeof (pte_t)) >PAGE_SIZE ); }
arch/arc/mm/tlb.c +2 -2
@@ -658,8 +658,8 @@ char *arc_mmu_mumbojumbo(int cpu_id, char *buf, int len) IS_USED_CFG(CONFIG_TRANSPARENT_HUGEPAGE)); n +=scnprintf (buf + n, len - n, -"MMU [v%x]\t: %dk PAGE, %sJTLB %d (%dx%d), uDTLB %d, uITLB %d%s%s\n" , - p_mmu->ver, p_mmu->pg_sz_k, super_pg, +"MMU [v%x]\t: %dk PAGE, %s, swalk %d lvl, JTLB %d (%dx%d), uDTLB %d, uITLB %d%s%s\n" , + p_mmu->ver, p_mmu->pg_sz_k, super_pg,CONFIG_PGTABLE_LEVELS , p_mmu->sets * p_mmu->ways, p_mmu->sets, p_mmu->ways, p_mmu->u_dtlb, p_mmu->u_itlb,IS_AVAIL2 (p_mmu->pae,", PAE40 " ,CONFIG_ARC_HAS_PAE40 ));
arch/arc/mm/tlbex.S +9 -0
@@ -173,6 +173,15 @@ ex_saved_reg1: tst r3, r3 bz do_slow_path_pf ; if no Page Table, do page fault +#if CONFIG_PGTABLE_LEVELS > 2 + lsr r0, r2, PMD_SHIFT ; Bits for indexing into PMD + and r0, r0, (PTRS_PER_PMD - 1) + ld.as r1, [r3, r0] ; PMD entry + tst r1, r1 + bz do_slow_path_pf + mov r3, r1 +#endif + #ifdef CONFIG_TRANSPARENT_HUGEPAGE and.f 0, r3, _PAGE_HW_SZ ; Is this Huge PMD (thp) add2.nz r1, r1, r0
Signature · 7 lines
2.25.1 _______________________________________________ linux-snps-arc mailing list [email protected] http://lists.infradead.org/mailman/listinfo/linux-snps-arc
Re: [PATCH v2 16/19] ARC: mm: support 3 levels of page tables
Quoted 32 lines · Vineet Gupta
ARCv2 MMU is software walked and Linux implements 2 levels of paging: pgd/pte. Forthcoming hw will have multiple levels, so this change preps mm code for same. It is also fun to try multi levels even on soft-walked code to ensure generic mm code is robust to handle.
overview
________2 levels {pgd, pte} : pmd is folded but pmd_* macros are valid and operate on pgd 3 levels {pgd, pmd, pte}:
- pud is folded and pud_* macros point to pgd
- pmd_* macros operate on actual pmdcode changes
____________1. #include <
asm-generic/pgtable-nopud.h>2. Define
CONFIG_PGTABLE_LEVELS33a. Define
PMD_SHIFT,PMD_SIZE,PMD_MASK,pmd_t3b. Definepmd_val()which actually deals with pmd
(pmd_offset(),pmd_index()are provided by generic code)
3c. pmd_alloc_one()/pmd_free() also provided by generic code
(pmd_populate/pmd_free already exist)4. Define
pud_none(),pud_bad()macros based on genericpud_val()which internally pertains to pgd now.
4b. definepud_populate()to just setup pgdSigned-off-by Vineet Gupta <[email protected]>---
...
Quoted 14 lines
arch/arc/include/asm/pgtable-levels.h +2
-0@@ -10,6 +10,8 @@#ifndef _ASM_ARC_PGTABLE_LEVELS_H#define _ASM_ARC_PGTABLE_LEVELS_H +#if CONFIG_PGTABLE_LEVELS ==2 +/* * 2 level paging setup for software walked MMUv3 (ARC700) and MMUv4 (HS) *
arch/arc/include/asm/pgtable-levels.h line 42
Quoted 6 lines
arch/arc/include/asm/pgtable-levels.h +1
-2@@ -37,16 +39,38 @@#define PGDIR_SHIFT 21#endif -#define PGDIR_SIZE BIT(PGDIR_SHIFT)/* vaddr span, not PDG sz */ -#define PGDIR_MASK (~(PGDIR_SIZE - 1)) +#else
Adding /* CONFIG_PGTABLE_LEVELS == 2 */ would make the whole thing a bit more readable, I think.
arch/arc/include/asm/pgtable-levels.h line 53
Quoted 11 lines
arch/arc/include/asm/pgtable-levels.h +11
-0+ +/* +* A default 3 level paging testing setup in software walked MMU +* MMUv4 (8K page): <4> : <7> : <8> : <13> +*/ +#define PGDIR_SHIFT 28 +#if CONFIG_PGTABLE_LEVELS >2 +#define PMD_SHIFT 21 +#endif + +#endif
and here as well.
Quoted 13 lines
arch/arc/include/asm/pgtable-levels.h +9
-1+#define PGDIR_SIZE BIT(PGDIR_SHIFT) +#define PGDIR_MASK (~(PGDIR_SIZE - 1))#define PTRS_PER_PGD BIT(32 - PGDIR_SHIFT) -#define PTRS_PER_PTE BIT(PGDIR_SHIFT - PAGE_SHIFT) +#if CONFIG_PGTABLE_LEVELS >2 +#define PMD_SIZE BIT(PMD_SHIFT) +#define PMD_MASK (~(PMD_SIZE - 1)) +#define PTRS_PER_PMD BIT(PGDIR_SHIFT - PMD_SHIFT) +#endif + +#define PTRS_PER_PTE BIT(PMD_SHIFT - PAGE_SHIFT)
Signature · 2 lines
Sincerely yours, Mike.
Vineet Gupta
maintainerRe: [PATCH v2 16/19] ARC: mm: support 3 levels of page tables
Quoted 52 lines · Mike Rapoport
Quoted 32 lines · Vineet Gupta
ARCv2 MMU is software walked and Linux implements 2 levels of paging: pgd/pte. Forthcoming hw will have multiple levels, so this change preps mm code for same. It is also fun to try multi levels even on soft-walked code to ensure generic mm code is robust to handle.
overview
________2 levels {pgd, pte} : pmd is folded but pmd_* macros are valid and operate on pgd 3 levels {pgd, pmd, pte}:
- pud is folded and pud_* macros point to pgd
- pmd_* macros operate on actual pmdcode changes
____________1. #include <
asm-generic/pgtable-nopud.h>2. Define
CONFIG_PGTABLE_LEVELS33a. Define
PMD_SHIFT,PMD_SIZE,PMD_MASK,pmd_t
3b. Definepmd_val()which actually deals with pmd
(pmd_offset(),pmd_index()are provided by generic code)
3c. pmd_alloc_one()/pmd_free() also provided by generic code
(pmd_populate/pmd_free already exist)4. Define
pud_none(),pud_bad()macros based on genericpud_val()which internally pertains to pgd now.
4b. definepud_populate()to just setup pgdSigned-off-by Vineet Gupta <[email protected]>---
...
Quoted 17 lines
arch/arc/include/asm/pgtable-levels.h +2
-0@@ -10,6 +10,8 @@#ifndef _ASM_ARC_PGTABLE_LEVELS_H#define _ASM_ARC_PGTABLE_LEVELS_H +#if CONFIG_PGTABLE_LEVELS ==2 +/* * 2 level paging setup for software walked MMUv3 (ARC700) and MMUv4 (HS) * @@ -37,16 +39,38 @@#define PGDIR_SHIFT 21#endif
Quoted 5 lines
arch/arc/include/asm/pgtable-levels.h line 42
Quoted 3 lines
arch/arc/include/asm/pgtable-levels.h +1
-2-#define PGDIR_SIZE BIT(PGDIR_SHIFT)/* vaddr span, not PDG sz */ -#define PGDIR_MASK (~(PGDIR_SIZE - 1)) +#else Adding /*
CONFIG_PGTABLE_LEVELS== 2 */ would make the whole thing a bit more readable, I think.
You meant
+#else /* CONFIG_PGTABLE_LEVELS != 2
Quoted 13 lines
arch/arc/include/asm/pgtable-levels.h line 53
Quoted 12 lines
arch/arc/include/asm/pgtable-levels.h +11
-0+ +/* +* A default 3 level paging testing setup in software walked MMU +* MMUv4 (8K page): <4> : <7> : <8> : <13> +*/ +#define PGDIR_SHIFT 28 +#if CONFIG_PGTABLE_LEVELS >2 +#define PMD_SHIFT 21 +#endif + +#endif and here as well.
I added following to indicate conditional coding for levels related code
+#endif /* CONFIG_PGTABLE_LEVELS */
Quoted 14 lines
Quoted 14 lines
arch/arc/include/asm/pgtable-levels.h +9
-1+#define PGDIR_SIZE BIT(PGDIR_SHIFT) +#define PGDIR_MASK (~(PGDIR_SIZE - 1))#define PTRS_PER_PGD BIT(32 - PGDIR_SHIFT) -#define PTRS_PER_PTE BIT(PGDIR_SHIFT - PAGE_SHIFT) +#if CONFIG_PGTABLE_LEVELS >2 +#define PMD_SIZE BIT(PMD_SHIFT) +#define PMD_MASK (~(PMD_SIZE - 1)) +#define PTRS_PER_PMD BIT(PGDIR_SHIFT - PMD_SHIFT) +#endif + +#define PTRS_PER_PTE BIT(PMD_SHIFT - PAGE_SHIFT)
Re: [PATCH v2 16/19] ARC: mm: support 3 levels of page tables
Quoted 57 lines · Vineet Gupta
Quoted 47 lines · Mike Rapoport
Quoted 32 lines · Vineet Gupta
ARCv2 MMU is software walked and Linux implements 2 levels of paging: pgd/pte. Forthcoming hw will have multiple levels, so this change preps mm code for same. It is also fun to try multi levels even on soft-walked code to ensure generic mm code is robust to handle.
overview
________2 levels {pgd, pte} : pmd is folded but pmd_* macros are valid and operate on pgd 3 levels {pgd, pmd, pte}:
- pud is folded and pud_* macros point to pgd
- pmd_* macros operate on actual pmdcode changes
____________1. #include <
asm-generic/pgtable-nopud.h>2. Define
CONFIG_PGTABLE_LEVELS33a. Define
PMD_SHIFT,PMD_SIZE,PMD_MASK,pmd_t
3b. Definepmd_val()which actually deals with pmd
(pmd_offset(),pmd_index()are provided by generic code)
3c. pmd_alloc_one()/pmd_free() also provided by generic code
(pmd_populate/pmd_free already exist)4. Define
pud_none(),pud_bad()macros based on genericpud_val()which internally pertains to pgd now.
4b. definepud_populate()to just setup pgdSigned-off-by Vineet Gupta <[email protected]>---
...
Quoted 12 lines
arch/arc/include/asm/pgtable-levels.h +2
-0@@ -10,6 +10,8 @@#ifndef _ASM_ARC_PGTABLE_LEVELS_H#define _ASM_ARC_PGTABLE_LEVELS_H +#if CONFIG_PGTABLE_LEVELS ==2 +/* * 2 level paging setup for software walked MMUv3 (ARC700) and MMUv4 (HS) * Quoted 8 lines
arch/arc/include/asm/pgtable-levels.h line 41
Quoted 6 lines
arch/arc/include/asm/pgtable-levels.h +1
-2@@ -37,16 +39,38 @@#define PGDIR_SHIFT 21#endif -#define PGDIR_SIZE BIT(PGDIR_SHIFT)/* vaddr span, not PDG sz */ -#define PGDIR_MASK (~(PGDIR_SIZE - 1)) +#else Adding /*
CONFIG_PGTABLE_LEVELS== 2 */ would make the whole thing a bit more readable, I think.
Quoted 3 lines
You meant
+#else /*
CONFIG_PGTABLE_LEVELS!= 2
I don't think we are consistent about it in the kernel, there are places that just copy the condition in #if and some change it to match the #else. I don't have a preference, up to you.
Quoted 31 lines
Quoted 13 lines
arch/arc/include/asm/pgtable-levels.h line 53
Quoted 12 lines
arch/arc/include/asm/pgtable-levels.h +11
-0+ +/* +* A default 3 level paging testing setup in software walked MMU +* MMUv4 (8K page): <4> : <7> : <8> : <13> +*/ +#define PGDIR_SHIFT 28 +#if CONFIG_PGTABLE_LEVELS >2 +#define PMD_SHIFT 21 +#endif + +#endif and here as well.
I added following to indicate conditional coding for levels related code
+#endif /*
CONFIG_PGTABLE_LEVELS*/Quoted 11 lines
Quoted 11 lines
arch/arc/include/asm/pgtable-levels.h +9
-1+#define PGDIR_SIZE BIT(PGDIR_SHIFT) +#define PGDIR_MASK (~(PGDIR_SIZE - 1))#define PTRS_PER_PGD BIT(32 - PGDIR_SHIFT) -#define PTRS_PER_PTE BIT(PGDIR_SHIFT - PAGE_SHIFT) +#if CONFIG_PGTABLE_LEVELS >2 +#define PMD_SIZE BIT(PMD_SHIFT) +#define PMD_MASK (~(PMD_SIZE - 1)) +#define PTRS_PER_PMD BIT(PGDIR_SHIFT - PMD_SHIFT) +#endif + +#define PTRS_PER_PTE BIT(PMD_SHIFT - PAGE_SHIFT)
Signature · 2 lines
Sincerely yours, Mike.
Vineet Gupta
maintainerPATCH v2 17/19 ARC: mm: support 4 levels of page tables
In reply to Vineet Gupta--- arch/arc/include/asm/page.h | 11 +++++++ arch/arc/include/asm/pgalloc.h | 11 +++++++ arch/arc/include/asm/pgtable-levels.h | 45 ++++++++++++++++++++++++--- arch/arc/mm/fault.c | 2 ++ arch/arc/mm/init.c | 1 + arch/arc/mm/tlbex.S | 9 ++++++ 6 files changed, 74 insertions(+), 5 deletions(-)
arch/arc/include/asm/page.h +11 -0
@@ -41,6 +41,17 @@ typedef struct {#define pgd_val (x) ((x).pgd)#define __pgd (x) ((pgd_t) { (x) }) +#if CONFIG_PGTABLE_LEVELS >3 + +typedef struct { +unsigned long pud; +}pud_t ; + +#define pud_val (x) ((x).pud) +#define __pud (x) ((pud_t) { (x) }) + +#endif + #if CONFIG_PGTABLE_LEVELS > 2 typedef struct {
arch/arc/include/asm/pgalloc.h +11 -0
@@ -70,6 +70,17 @@ static inline pgd_t *pgd_alloc(struct mm_struct *mm)return ret; } +#if CONFIG_PGTABLE_LEVELS >3 + +static inline void p4d_populate (struct mm_struct *mm,p4d_t *p4dp,pud_t *pudp) +{ +set_p4d (p4dp,__p4d ((unsigned long )pudp)); +} + +#define __pud_free_tlb (tlb, pmd, addr) pud_free((tlb)->mm, pmd) + +#endif +#if CONFIG_PGTABLE_LEVELS >2 static inline void pud_populate (struct mm_struct *mm,pud_t *pudp,pmd_t *pmdp)
arch/arc/include/asm/pgtable-levels.h +40 -5
88 lines not on this page · raw message
arch/arc/mm/fault.c +2 -0
@@ -46,6 +46,8 @@ noinline static int handle_kernel_vaddr_fault(unsigned long address)if (!p4d_present (*p4d_k)) goto bad_area; +set_p4d (p4d, *p4d_k); + pud =pud_offset (p4d, address); pud_k =pud_offset (p4d_k, address);if (!pud_present (*pud_k))
arch/arc/mm/init.c +1 -0
@@ -191,6 +191,7 @@ void __init mem_init(void)highmem_init ();BUILD_BUG_ON ((PTRS_PER_PGD *sizeof (pgd_t)) >PAGE_SIZE ); +BUILD_BUG_ON ((PTRS_PER_PUD *sizeof (pud_t)) >PAGE_SIZE );BUILD_BUG_ON ((PTRS_PER_PMD *sizeof (pmd_t)) >PAGE_SIZE );BUILD_BUG_ON ((PTRS_PER_PTE *sizeof (pte_t)) >PAGE_SIZE ); }
arch/arc/mm/tlbex.S +9 -0
@@ -173,6 +173,15 @@ ex_saved_reg1: tst r3, r3 bz do_slow_path_pf ; if no Page Table, do page fault +#if CONFIG_PGTABLE_LEVELS > 3 + lsr r0, r2, PUD_SHIFT ; Bits for indexing into PUD + and r0, r0, (PTRS_PER_PUD - 1) + ld.as r1, [r3, r0] ; PMD entry + tst r1, r1 + bz do_slow_path_pf + mov r3, r1 +#endif + #if CONFIG_PGTABLE_LEVELS > 2 lsr r0, r2, PMD_SHIFT ; Bits for indexing into PMD and r0, r0, (PTRS_PER_PMD - 1)
Signature · 7 lines
2.25.1 _______________________________________________ linux-snps-arc mailing list [email protected] http://lists.infradead.org/mailman/listinfo/linux-snps-arc
Vineet Gupta
maintainerPATCH v2 18/19 ARC: mm: vmalloc sync from kernel to user table to update PMD ...
In reply to Vineet Gupta... not PGD
vmalloc() sets up the kernel page table (starting from @swapper_pg_dir). But when vmalloc area is accessed in context of a user task, say opening terminal in n_tty_open(), the user page tables need to be synced from kernel page tables so that TLB entry is created in "user context".
The old code was doing this incorrectly, as it was updating the user pgd entry (first level itself) to point to kernel pud table (2nd level), effectively yanking away the entire user space translation with kernel one.
The correct way to do this is to ONLY update a user space pgd/pud/pmd entry if it is not popluated already. This ensures that only the missing leaf pmd entry gets updated to point to relevant kernel pte table.
From code change pov, we are chaging the pattern:
p4d = p4d_offset(pgd, address); p4d_k = p4d_offset(pgd_k, address); if (!p4d_present(*p4d_k)) goto bad_area; set_p4d(p4d, *p4d_k);
with
p4d = p4d_offset(pgd, address); p4d_k = p4d_offset(pgd_k, address); if (p4d_none(*p4d_k)) goto bad_area; if (!p4d_present(*p4d)) set_p4d(p4d, *p4d_k);
--- arch/arc/mm/fault.c | 24 ++++++++++++------------ 1 file changed, 12 insertions(+), 12 deletions(-)
arch/arc/mm/fault.c +12 -12
@@ -36,31 +36,31 @@ noinline static int handle_kernel_vaddr_fault(unsigned long address) pgd =pgd_offset (current->active_mm, address); pgd_k =pgd_offset_k (address); -if (!pgd_present (*pgd_k)) +if (pgd_none (*pgd_k)) goto bad_area; - -set_pgd (pgd, *pgd_k); +if (!pgd_present (*pgd)) +set_pgd (pgd, *pgd_k); p4d =p4d_offset (pgd, address); p4d_k =p4d_offset (pgd_k, address); -if (!p4d_present (*p4d_k)) +if (p4d_none (*p4d_k)) goto bad_area; - -set_p4d (p4d, *p4d_k); +if (!p4d_present (*p4d)) +set_p4d (p4d, *p4d_k); pud =pud_offset (p4d, address); pud_k =pud_offset (p4d_k, address); -if (!pud_present (*pud_k)) +if (pud_none (*pud_k)) goto bad_area; - -set_pud (pud, *pud_k); +if (!pud_present (*pud)) +set_pud (pud, *pud_k); pmd =pmd_offset (pud, address); pmd_k =pmd_offset (pud_k, address); -if (!pmd_present (*pmd_k)) +if (pmd_none (*pmd_k)) goto bad_area; - -set_pmd (pmd, *pmd_k); +if (!pmd_present (*pmd)) +set_pmd (pmd, *pmd_k);/* XXX: create the TLB entry here */ return 0 ;
Signature · 7 lines
2.25.1 _______________________________________________ linux-snps-arc mailing list [email protected] http://lists.infradead.org/mailman/listinfo/linux-snps-arc
Vineet Gupta
maintainerPATCH v2 19/19 ARC: mm: introduce _PAGE_TABLE to explicitly link pgd,pud,pmd entries
In reply to Vineet GuptaARCv3 hardware walker expects Table Descriptors to have b'11 in LSB bits to continue moving to next level.
This commits adds that (to ARCv2 code) and ensures that it works in software walked regime.
The pte entries stil need tagging, but that is not possible in ARCv2 since the LSB 2 bits are currently used.
--- arch/arc/include/asm/pgalloc.h | 8 ++++---- arch/arc/include/asm/pgtable-bits-arcv2.h | 2 ++ arch/arc/include/asm/pgtable-levels.h | 6 +++--- arch/arc/mm/tlbex.S | 4 +++- 4 files changed, 12 insertions(+), 8 deletions(-)
arch/arc/include/asm/pgalloc.h +4 -4
@@ -43,12 +43,12 @@ pmd_populate_kernel(struct mm_struct *mm, pmd_t *pmd, pte_t *pte)* * The cast itself is needed given simplistic definition of set_pmd() */ -set_pmd (pmd,__pmd ((unsigned long )pte)); +set_pmd (pmd,__pmd ((unsigned long )pte | _PAGE_TABLE)); }static inline void pmd_populate (struct mm_struct *mm,pmd_t *pmd,pgtable_t pte_page) { -set_pmd (pmd,__pmd ((unsigned long )page_address (pte_page))); +set_pmd (pmd,__pmd ((unsigned long )page_address (pte_page) | _PAGE_TABLE)); }static inline pgd_t *pgd_alloc (struct mm_struct *mm)
@@ -74,7 +74,7 @@ static inline pgd_t *pgd_alloc(struct mm_struct *mm)static inline void p4d_populate (struct mm_struct *mm,p4d_t *p4dp,pud_t *pudp) { -set_p4d (p4dp,__p4d ((unsigned long )pudp)); +set_p4d (p4dp,__p4d ((unsigned long )pudp | _PAGE_TABLE)); }#define __pud_free_tlb (tlb, pmd, addr) pud_free((tlb)->mm, pmd)
@@ -85,7 +85,7 @@ static inline void p4d_populate(struct mm_struct *mm, p4d_t *p4dp, pud_t *pudp)static inline void pud_populate (struct mm_struct *mm,pud_t *pudp,pmd_t *pmdp) { -set_pud (pudp,__pud ((unsigned long )pmdp)); +set_pud (pudp,__pud ((unsigned long )pmdp | _PAGE_TABLE)); }#define __pmd_free_tlb (tlb, pmd, addr) pmd_free((tlb)->mm, pmd)
arch/arc/include/asm/pgtable-bits-arcv2.h +2 -0
@@ -32,6 +32,8 @@#define _PAGE_HW_SZ 0#endif +#define _PAGE_TABLE 0x3 +/* Defaults for every user page */ #define ___DEF (_PAGE_PRESENT | _PAGE_CACHEABLE)
arch/arc/include/asm/pgtable-levels.h +3 -3
@@ -98,7 +98,7 @@/* In 4 level paging, p4d_* macros work on pgd */ #define p4d_none (x) (!p4d_val(x)) -#define p4d_bad (x) ((p4d_val(x) & ~PAGE_MASK)) +#define p4d_bad (x) (!(p4d_val(x) & _PAGE_TABLE))#define p4d_present (x) (p4d_val(x))#define p4d_clear (xp) do { p4d_val(*(xp)) = 0; } while (0)#define p4d_pgtable (p4d) ((pud_t *)(p4d_val(p4d) & PAGE_MASK))
@@ -120,7 +120,7 @@* In 4 level paging, pud_* macros work on pud */ #define pud_none (x) (!pud_val(x)) -#define pud_bad (x) ((pud_val(x) & ~PAGE_MASK)) +#define pud_bad (x) (!(pud_val(x) & _PAGE_TABLE))#define pud_present (x) (pud_val(x))#define pud_clear (xp) do { pud_val(*(xp)) = 0; } while (0)#define pud_pgtable (pud) ((pmd_t *)(pud_val(pud) & PAGE_MASK))
@@ -147,7 +147,7 @@* In 3+ level paging (pgd -> pmd -> pte), pmd_* macros work on pmd */ #define pmd_none (x) (!pmd_val(x)) -#define pmd_bad (x) ((pmd_val(x) & ~PAGE_MASK)) +#define pmd_bad (pmd) (!(pmd_val(pmd) & _PAGE_TABLE))#define pmd_present (x) (pmd_val(x))#define pmd_clear (xp) do { pmd_val(*(xp)) = 0; } while (0)#define pmd_page_vaddr (pmd) (pmd_val(pmd) & PAGE_MASK)
arch/arc/mm/tlbex.S +3 -1
@@ -171,11 +171,12 @@ ex_saved_reg1: lsr r0, r2, PGDIR_SHIFT ; Bits for indexing into PGD ld.as r3, [r1, r0] ; PGD entry corresp to faulting addr tst r3, r3 - bz do_slow_path_pf ; if no Page Table, do page fault + bz do_slow_path_pf ; next level table missing, handover to linux vm code #if CONFIG_PGTABLE_LEVELS > 3 lsr r0, r2, PUD_SHIFT ; Bits for indexing into PUD and r0, r0, (PTRS_PER_PUD - 1) + bmskn r3, r3, 1 ; clear _PAGE_TABLE bits ld.as r1, [r3, r0] ; PMD entry tst r1, r1 bz do_slow_path_pf
@@ -185,6 +186,7 @@ ex_saved_reg1: #if CONFIG_PGTABLE_LEVELS > 2 lsr r0, r2, PMD_SHIFT ; Bits for indexing into PMD and r0, r0, (PTRS_PER_PMD - 1) + bmskn r3, r3, 1 ; clear _PAGE_TABLE bits ld.as r1, [r3, r0] ; PMD entry tst r1, r1 bz do_slow_path_pf
Signature · 7 lines
2.25.1 _______________________________________________ linux-snps-arc mailing list [email protected] http://lists.infradead.org/mailman/listinfo/linux-snps-arc
Quoted 6 lines · Vineet Gupta
Hi,
Big pile of ARC mm changes to prepare for 3 or 4 levels of paging (from current 2) needed for new hardware page walked MMUv6 (in aRCv3 ISA based cores).
Quoted 6 lines
Most of these changes are incremental cleanups to make way for 14/18 and 15/18 which actually imeplement the new levels (in existing ARCv2 port) and worth a critical eye.
CC'ing some of you guys dealing with page tables for a while :-) to spot any obvious gotchas.
There are a couple of small nits here and there, but overall
Quoted 62 lines
Thx,
-VineetChanges since v1 [1]
- Switched ARC toasm-generic/pgalloc.h(so struct page basedpgtable_t) [Mike Rapoport]
- Dropped {pud,pmd}_alloc_one/{pud,pmd}_free provided byasm-generic/pgalloc.h[Mike Rapoport]
- Negative diffstat now due to above
- AddedBUILD_BUG_ON()toarch/arc/mm/init.cfor sanity of table sizes
- Consolidated 2 patches related toARC_USE_SCRATCH_REG[Mike Rapoport]
- Reworked how mmu is re-enabled in entry code [Jose Abreu][1] http://lists.infradead.org/pipermail/linux-snps-arc/2021-August/005326.html
Vineet Gupta (19):
ARC: mm: useSCRATCH_DATA0register for caching pgdir in ARCv2 only
ARC: mm: remove tlb paranoid code
ARC: mm: move mmu/cache externs out tosetup.h
ARC: mm: Fixes to allowSTRICT_MM_TYPECHECKS
ARC: mm: EnableSTRICT_MM_TYPECHECKS
ARC: ioremap: use more commonly usedPAGE_KERNELbased uncached flag
ARC: mm: pmd_populate* to use the canonicalset_pmd(and droppmd_set)
ARC: mm: switchpgtable_tback to struct page *
ARC: mm: switch toasm-generic/pgalloc.h
ARC: mm: non-functional code cleanup ahead of 3 levels
ARC: mm: move MMU specific bits out of ASID allocator
ARC: mm: move MMU specific bits out of entry code ...
ARC: mm: disintegratemmu.h(arcv2 bits out)
ARC: mm: disintegratepgtable.hinto levels and flags
ARC: mm: hack to allow 2 level build with 4 level code
ARC: mm: support 3 levels of page tables
ARC: mm: support 4 levels of page tables
ARC: mm: vmalloc sync from kernel to user table to update PMD ...
ARC: mm: introduce_PAGE_TABLEto explicitly link pgd,pud,pmd entriesarch/arc/Kconfig | 7 +- arch/arc/include/asm/cache.h | 4 - arch/arc/include/asm/entry-compact.h | 8 - arch/arc/include/asm/mmu-arcv2.h | 103 +++++++ arch/arc/include/asm/mmu.h | 73 +---- arch/arc/include/asm/mmu_context.h | 28 +- arch/arc/include/asm/page.h | 74 +++-- arch/arc/include/asm/pgalloc.h | 81 ++---- arch/arc/include/asm/pgtable-bits-arcv2.h | 151 +++++++++++ arch/arc/include/asm/pgtable-levels.h | 179 ++++++++++++ arch/arc/include/asm/pgtable.h | 315 +--------------------- arch/arc/include/asm/processor.h | 2 +- arch/arc/include/asm/setup.h | 12 +- arch/arc/kernel/entry-arcv2.S | 1 + arch/arc/kernel/entry.S | 7 +- arch/arc/mm/fault.c | 20 +- arch/arc/mm/init.c | 5 + arch/arc/mm/ioremap.c | 3 +- arch/arc/mm/tlb.c | 68 +---- arch/arc/mm/tlbex.S | 78 ++---- 20 files changed, 591 insertions(+), 628 deletions(-) create mode 100644 arch/arc/include/asm/mmu-arcv2.h create mode 100644 arch/arc/include/asm/pgtable-bits-arcv2.h create mode 100644 arch/arc/include/asm/pgtable-levels.hSignature · 1 line
2.25.1
Signature · 2 lines
Sincerely yours, Mike.
Vineet Gupta
maintainerQuoted 7 lines · Mike Rapoport
Quoted 6 lines · Vineet Gupta
Hi,
Big pile of ARC mm changes to prepare for 3 or 4 levels of paging (from current 2) needed for new hardware page walked MMUv6 (in aRCv3 ISA based cores).
Quoted 7 lines
Quoted 6 lines
Most of these changes are incremental cleanups to make way for 14/18 and 15/18 which actually imeplement the new levels (in existing ARCv2 port) and worth a critical eye.
CC'ing some of you guys dealing with page tables for a while :-) to spot any obvious gotchas.
There are a couple of small nits here and there, but overall
I've fixed the last remaining things locally and won't repost, unless you want me to.
Quoted 2 lines
Acked-by Mike Rapoport <[email protected]>
Thx for spending time to review this Mike. Much appreciated.
-Vineet
Quoted 63 lines
Quoted 63 lines
Thx,
-VineetChanges since v1 [1]
- Switched ARC toasm-generic/pgalloc.h(so struct page basedpgtable_t) [Mike Rapoport]
- Dropped {pud,pmd}_alloc_one/{pud,pmd}_free provided byasm-generic/pgalloc.h[Mike Rapoport]
- Negative diffstat now due to above
- AddedBUILD_BUG_ON()toarch/arc/mm/init.cfor sanity of table sizes
- Consolidated 2 patches related toARC_USE_SCRATCH_REG[Mike Rapoport]
- Reworked how mmu is re-enabled in entry code [Jose Abreu][1] http://lists.infradead.org/pipermail/linux-snps-arc/2021-August/005326.html
Vineet Gupta (19):
ARC: mm: useSCRATCH_DATA0register for caching pgdir in ARCv2 only
ARC: mm: remove tlb paranoid code
ARC: mm: move mmu/cache externs out tosetup.h
ARC: mm: Fixes to allowSTRICT_MM_TYPECHECKS
ARC: mm: EnableSTRICT_MM_TYPECHECKS
ARC: ioremap: use more commonly usedPAGE_KERNELbased uncached flag
ARC: mm: pmd_populate* to use the canonicalset_pmd(and droppmd_set)
ARC: mm: switchpgtable_tback to struct page *
ARC: mm: switch toasm-generic/pgalloc.h
ARC: mm: non-functional code cleanup ahead of 3 levels
ARC: mm: move MMU specific bits out of ASID allocator
ARC: mm: move MMU specific bits out of entry code ...
ARC: mm: disintegratemmu.h(arcv2 bits out)
ARC: mm: disintegratepgtable.hinto levels and flags
ARC: mm: hack to allow 2 level build with 4 level code
ARC: mm: support 3 levels of page tables
ARC: mm: support 4 levels of page tables
ARC: mm: vmalloc sync from kernel to user table to update PMD ...
ARC: mm: introduce_PAGE_TABLEto explicitly link pgd,pud,pmd entriesarch/arc/Kconfig | 7 +- arch/arc/include/asm/cache.h | 4 - arch/arc/include/asm/entry-compact.h | 8 - arch/arc/include/asm/mmu-arcv2.h | 103 +++++++ arch/arc/include/asm/mmu.h | 73 +---- arch/arc/include/asm/mmu_context.h | 28 +- arch/arc/include/asm/page.h | 74 +++-- arch/arc/include/asm/pgalloc.h | 81 ++---- arch/arc/include/asm/pgtable-bits-arcv2.h | 151 +++++++++++ arch/arc/include/asm/pgtable-levels.h | 179 ++++++++++++ arch/arc/include/asm/pgtable.h | 315 +--------------------- arch/arc/include/asm/processor.h | 2 +- arch/arc/include/asm/setup.h | 12 +- arch/arc/kernel/entry-arcv2.S | 1 + arch/arc/kernel/entry.S | 7 +- arch/arc/mm/fault.c | 20 +- arch/arc/mm/init.c | 5 + arch/arc/mm/ioremap.c | 3 +- arch/arc/mm/tlb.c | 68 +---- arch/arc/mm/tlbex.S | 78 ++---- 20 files changed, 591 insertions(+), 628 deletions(-)create mode 100644
arch/arc/include/asm/mmu-arcv2.hcreate mode 100644arch/arc/include/asm/pgtable-bits-arcv2.hcreate mode 100644arch/arc/include/asm/pgtable-levels.hSignature · 1 line
2.25.1
Quoted 12 lines · Vineet Gupta
Quoted 11 lines · Mike Rapoport
Quoted 10 lines · Vineet Gupta
Hi,
Big pile of ARC mm changes to prepare for 3 or 4 levels of paging (from current 2) needed for new hardware page walked MMUv6 (in aRCv3 ISA based cores).
Most of these changes are incremental cleanups to make way for 14/18 and 15/18 which actually imeplement the new levels (in existing ARCv2 port) and worth a critical eye.
Quoted 6 lines
Quoted 3 lines
Quoted 2 lines
CC'ing some of you guys dealing with page tables for a while :-) to spot any obvious gotchas.
There are a couple of small nits here and there, but overall
I've fixed the last remaining things locally and won't repost, unless you want me to.
No need.
Quoted 4 lines
Quoted 2 lines
Acked-by Mike Rapoport <[email protected]>Thx for spending time to review this Mike. Much appreciated.
You are welcome :)
Quoted 67 lines
-Vineet
Quoted 63 lines
Quoted 63 lines
Thx,
-VineetChanges since v1 [1]
- Switched ARC toasm-generic/pgalloc.h(so struct page basedpgtable_t) [Mike Rapoport]
- Dropped {pud,pmd}_alloc_one/{pud,pmd}_free provided byasm-generic/pgalloc.h[Mike Rapoport]
- Negative diffstat now due to above
- AddedBUILD_BUG_ON()toarch/arc/mm/init.cfor sanity of table sizes
- Consolidated 2 patches related toARC_USE_SCRATCH_REG[Mike Rapoport]
- Reworked how mmu is re-enabled in entry code [Jose Abreu][1] http://lists.infradead.org/pipermail/linux-snps-arc/2021-August/005326.html
Vineet Gupta (19):
ARC: mm: useSCRATCH_DATA0register for caching pgdir in ARCv2 only
ARC: mm: remove tlb paranoid code
ARC: mm: move mmu/cache externs out tosetup.h
ARC: mm: Fixes to allowSTRICT_MM_TYPECHECKS
ARC: mm: EnableSTRICT_MM_TYPECHECKS
ARC: ioremap: use more commonly usedPAGE_KERNELbased uncached flag
ARC: mm: pmd_populate* to use the canonicalset_pmd(and droppmd_set)
ARC: mm: switchpgtable_tback to struct page *
ARC: mm: switch toasm-generic/pgalloc.h
ARC: mm: non-functional code cleanup ahead of 3 levels
ARC: mm: move MMU specific bits out of ASID allocator
ARC: mm: move MMU specific bits out of entry code ...
ARC: mm: disintegratemmu.h(arcv2 bits out)
ARC: mm: disintegratepgtable.hinto levels and flags
ARC: mm: hack to allow 2 level build with 4 level code
ARC: mm: support 3 levels of page tables
ARC: mm: support 4 levels of page tables
ARC: mm: vmalloc sync from kernel to user table to update PMD ...
ARC: mm: introduce_PAGE_TABLEto explicitly link pgd,pud,pmd entriesarch/arc/Kconfig | 7 +- arch/arc/include/asm/cache.h | 4 - arch/arc/include/asm/entry-compact.h | 8 - arch/arc/include/asm/mmu-arcv2.h | 103 +++++++ arch/arc/include/asm/mmu.h | 73 +---- arch/arc/include/asm/mmu_context.h | 28 +- arch/arc/include/asm/page.h | 74 +++-- arch/arc/include/asm/pgalloc.h | 81 ++---- arch/arc/include/asm/pgtable-bits-arcv2.h | 151 +++++++++++ arch/arc/include/asm/pgtable-levels.h | 179 ++++++++++++ arch/arc/include/asm/pgtable.h | 315 +--------------------- arch/arc/include/asm/processor.h | 2 +- arch/arc/include/asm/setup.h | 12 +- arch/arc/kernel/entry-arcv2.S | 1 + arch/arc/kernel/entry.S | 7 +- arch/arc/mm/fault.c | 20 +- arch/arc/mm/init.c | 5 + arch/arc/mm/ioremap.c | 3 +- arch/arc/mm/tlb.c | 68 +---- arch/arc/mm/tlbex.S | 78 ++---- 20 files changed, 591 insertions(+), 628 deletions(-)create mode 100644
arch/arc/include/asm/mmu-arcv2.hcreate mode 100644arch/arc/include/asm/pgtable-bits-arcv2.hcreate mode 100644arch/arc/include/asm/pgtable-levels.hSignature · 1 line
2.25.1
Signature · 2 lines
Sincerely yours, Mike.